Submicron Aerosol Composition and Source Contribution across the Kathmandu Valley, Nepal, in Winter

Abstract

The Kathmandu valley experiences an average wintertime PM₁ concentration of ∼100 μg m^–3 and daily peaks over 200 μg m^–3. We present ambient nonrefractory PM₁ chemical composition, and concentration measured by a mini aerosol mass spectrometer (mAMS) sequentially at Dhulikhel (on the valley exterior), then urban Ratnapark, and finally suburban Lalitpur in winter 2018. At all sites, organic aerosol (OA) was the largest contributor to combined PM₁ (C-PM₁) (49%) and black carbon (BC) was the second largest contributor (21%). The average background C-PM₁ at Dhulikhel was 48 μg m^–3; the urban enhancement was 120% (58 μg m^–3). BC had an average of 6.1 μg m^–3 at Dhulikhel, an urban enhancement of 17.4 μg m^–3. Sulfate (SO₄) was 3.6 μg m^–3 at Dhulikhel, then 7.5 μg m^–3 at Ratnapark, and 12.0 μg m^–3 at Lalitpur in the brick kiln region. Chloride (Chl) increased by 330 and 250% from Dhulikhel to Ratnapark and Lalitpur on average. Positive matrix factorization (PMF) identified seven OA sources, four primary OA sources, hydrocarbon-like (HOA), biomass burning (BBOA), trash burning (TBOA), a sulfate-containing local OA source (sLOA), and three secondary oxygenated organic aerosols (OOA). OOA was the largest fraction of OA, over 50% outside the valley and 36% within. HOA (traffic) was the most prominent primary source, contributing 21% of all OA and 44% of BC. Brick kilns were the second largest contributor to C-PM₁, 12% of OA, 33% of BC, and a primary emitter of aerosol sulfate. These results, though successive, indicate the importance of multisite measurements to understand ambient particulate matter concentration heterogeneity across urban regions.

Introduction

Unhealthy urban air is a global problem responsible for ∼3.75 million premature deaths every year.¹ Degraded air quality (AQ) significantly decreases life expectancy, especially in the global South.^2,3 In South Asia, poor air quality with especially severe wintertime pollution exceeds the Nepali, Indian, and the World Health Organization (WHO) PM_2.5 standards.⁴⁻⁶ The region has seasonal cycles, where monsoon has minimal PM₁₀ concentrations and winter has the highest.⁷ Nepal is located along the northern edge of the Indo-Gangetic plain (IGP), and within the Himalayan Mountains which contains the densely populated Kathmandu Valley. This urbanized basin has a range of topography, population density, industrial usage, and meteorology. The downtown region is the country’s population center, with an international airport and high vehicle traffic levels. The valley’s rural edges often rely on locally available energy sources (e.g., combustion of wood, dung, other biomass, or coal). A patchwork of suburban towns, industrial zones, and agricultural activity lies between these extremes creating spatial differences in air quality throughout the valley.

Mountain ridges surrounding the Kathmandu Valley reduce the impact of winds that otherwise would transport pollutants from the IGP or disperse local emissions.⁸ Overnight temperature inversions suppress vertical mixing, driving pollutant concentrations to dangerous levels. Increased mixed-layer depths and airflow through mountain and river passes in the daytime partially reduce air pollution levels by dilution and transport out of the valley.⁹ A continual population increase has compounded the air quality problem caused by the basin topography. The population increased by almost one million persons (65%) from 2001 to 2011.^10,11 Population growth increased energy demand and fuel consumption by 3.2% a year from 2005 to 2010.¹²

The previous air quality campaigns including the initial Nepal Ambient Monitoring and Source Testing Experiment (NAMaSTE) established the primary regional sources of particulate matter (PM) as vehicle emissions, coal combustion, biomass burning (BB) for cooking and agriculture, trash burning (TB), dust, and industry.¹³⁻²⁰ Most PM_2.5 in Kathmandu Valley is locally generated, though wind introduces some fraction of pollutants from the IGP.^21,22 Elevated black carbon (BC) levels are typical in the region.²³⁻²⁵ Previous studies such as sustainable atmosphere for the Kathmandu Valley (SusKat) have measured gas-phase and PM₁₀ compositions, noted the effects of primary and secondary sources on air quality in the region,^16,26−29 and apportioned pre-monsoon PM measurements to their sources.^18,20,30,31 However, wintertime submicron aerosol composition and the spatial impact of individual point sources on ambient concentrations remained vague.

From 2000 to 2018, the vehicle fleet increased by 14% per year, over one million cars by 2010. The Bagmati Zone, containing Kathmandu, has half the country’s vehicles. In Nepal, the EURO VI emission standard is required for new vehicles, a quarter of vehicles use the outdated EURO II standard.³² Emissions from the vehicle fleet, likely driven by these older vehicles, are the dominant regional PM source.^19,20 The Bharat Stage VI, a EURO 4 equivalent standard for fuel, was adopted by India in 2017, limiting the sulfur content in diesel fuel from 350 to 50 ppm. Nepal imports nearly all its petroleum from India and therefore uses the same EURO VI fuel standards. The EURO VI standard is less stringent than that used in Europe; the sulfur limit is 5 times higher. This high level of sulfur in fuel implicates diesel vehicles contributing to sulfate aerosol.^32,33

Agricultural waste burning significantly emits aerosol, specifically organic aerosol (OA) and chloride.³⁴ Source testing from NAMaSTE in 2015 found that biomass fueled cooking, chlorinated plastics, and agricultural residue burning experiments emitted large fractions of Chl.³⁵ The routine use of these fires, their significant emission rate, and the high fraction of Chl in those emissions indicate the importance of crop and garbage burning as large atmospheric chloride sources.

Atmospheric chlorides have natural sources, (e.g., wildfires, dust storms, and volcanoes); however, seawater is the primary source.^36,37 Seawater emissions of Chl are typically 100 times higher than other sources such as dust.³⁸ As Nepal is landlocked, marine emissions are likely minimal. The known anthropogenic emission in South Asia, are coal combustion, open burning of garbage, trash, or grass, and residential use of dung or hardwood fuels.^19,20,35,39 Other observations in South Asia have noted high chloride levels in submicron aerosol.⁴⁰ Gas-phase HCl emissions from the same sources are similarly high.^38,41

This campaign provides Kathmandu’s first real-time, wintertime measurements of submicron aerosol chemical composition characterizing the spatial and temporal variation of PM₁, other pollutant concentrations, and air quality (AQ) via sequential monitoring at multiple sites across the Kathmandu Valley. We quantify the differences between urban, suburban, and rural regions and how nearby land use influences air pollution at the basin scale. We show how emissions, topography, and meteorology drive the daily patterns, chemical composition, and temporal behavior of PM₁ at each of the three successive sites. Positive matrix factorization (PMF) is used to understand the source-specific contribution of OA to ambient air quality. Our observations can provide quantitative data to inform air quality strategies for municipalities throughout South Asia.

Experimental Section

General Methods

The NAMaSTE campaigns focused in-depth on the air quality of Nepal’s largest urban area, the Kathmandu Valley. In April 2015, NAMaSTE1 measured the emissions of major South Asian combustion sources in unprecedented detail and the ambient pre-monsoon concentrations in suburban Bode, Madhyapur, in the Kathmandu Valley (Figure 1^{19,20,35,41,42}). The Ghorka earthquake on April 25, 2015, prematurely ended the initial ambient campaign. This work significantly expanded the scope of the ambient measurements by utilizing three Kathmandu Valley sites, Dhulikhel, Ratnapark, and Lalitpur, and probing a different season during winter (January–February) of 2018 (Figure 1). The multiple measurement sites established a gradient across land use, population, elevation, and geographic location. The wintertime 2018 measurements are compared with the initial pre-monsoon deployment in April 2015 providing additional spatial and seasonal context and a more comprehensive understanding of the year-round air pollution issues. The measurement sites established a gradient across land use and population, elevation, and geographic location. The NAMaSTE2 campaign measured different urban densities in the valley interior, downtown (Ratnapark), peri-urban (Lalitpur), and rural along the valley rim at (Dhulikhel). Each site was removed from individual emission sources direct plumes while reflecting common local sources and use.

Figure 1.

Map of monitoring site locations for NAMaSTE2, Dhulikhel, Ratna Park, and Lalitpur, as well as NAMaSTE1 site of Bode, and inset of Nepal from Open Street Map.

These measurements are taken within a 6-week period, but their sequential nature limits quantitative comparisons. Nonetheless, the measurements probed typical wintertime meteorology and conditions at each site and thus the comparisons and differences observed are informative.^7−9,16,28

Site Selection and Descriptions

The first measurement site, Dhulikhel (27.608°N, 85.547°E, 1650 m.a.s.l), was located at the Kathmandu University School of Medical Sciences, along the external rim of the Kathmandu Valley (January 5th, 2018, to January 14th, 2018). This site’s rural nature and reduced local emissions represent a reasonable regional background. Anabatic and katabatic flow dominated the meteorology of this site. The meteorology of all sites is detailed below.

The second measurement site, Ratnapark (27.707°N, 85.315°E, 1500 m.a.s.l.), was in downtown Kathmandu, on the valley floor representing the valley’s urban center (January 18th and January 29th, 2018). This site sits approximately 35 m from the road in a city park, near a congested intersection of the main avenue to the downtown area. The street was lined by numerous outdoor markets and food vendors.

The NAMaSTE2 campaign’s ambient measurements at Dhulikhel and Ratnapark sites were made at preexisting joint International Center for Integrated Mountain Development (ICIMOD) and Nepali Department of Environment (DoEnv) air quality measurement sites within the valley. These are two in a series of converted shipping containers that monitor air pollution in the Hindu Kush range.

The third site, Lalitpur (27.646°N, 85.322°E, 1525 m.a.s.l), was on the third floor of the ICIMOD headquarters in suburbs near the south rim of the valley (January 30th to February 10th, 2018). This site is in a primarily residential area but near several brick kilns, and other industries. It is removed from the urban center and shares many characteristics with the Bode site in the NAMaSTE1 and SusKat experiments.

Aerosol Sampling Instrumentation and Analysis

The mini aerosol mass spectrometer (mAMS; Aerodyne Research, Inc.) measures real-time nonrefractory submicron aerosol mass concentration and the size-resolved chemical composition of species that volatilize at 600 °C under vacuum (nonrefractory PM₁; NR-PM₁). NR-PM₁ includes the major aerosol components: organic aerosol, sulfate, nitrate, chloride, and ammonium. The mAMS is a version of Aerodyne’s compact time-of-flight mass spectrometer⁴³ with a smaller vacuum chamber and the pumping system from a time-of-flight aerosol chemical speciation monitor (TOF-ACSM).⁴⁴ The source measurements made in 2015 used the same mAMS.³⁵

The mAMS measured size-resolved chemical composition using the efficient particle time-of-flight (ePToF) chopper system an improved system over traditional PTOF data.^45,46 The mAMS sampled 10 s of mass spectrum (MS) then 30 s of ePToF mode for 1 h followed by 5 min of single-particle (SP) mode. Interpreting SP mode measurements is challenging and beyond the scope of this campaign. All samples measure mass spectra from m/z 10 to 300, with the vaporizer set to 600 °C. All ambient measurements assume the literature standard collection efficiency (CE) of 0.5.^47,48 Based on previous cross instrument comparisons, the mAMS CE is assumed to be constant for the experiment’s duration (e.g., refs (43) and (49−52)). We performed lens alignment, particle size, and ionization efficiency (IE) calibrations using aerosolized ammonium nitrate. The mAMS calibrations for IE used size-selected ammonia nitrate (e.g., 604.3, 453, 406.8, 305.1, and 60.4 nm, detailed in Table S1). The entire mAMS dataset presented uses a campaign average IE value normalized for airbeam (IE/AB) ((3.25 × 10^–14) ± (3 × 10^–15)). The mAMS detection limit was measured daily with 30 min of sampling through a clean, high-efficiency particulate air (HEPA) filter. The detection limit for each species is defined as three standard deviations (σ) of the combined filter periods.⁵³ The detection limits for the mAMS aerosol species at each site are in Table S2. The assumed overall uncertainty for total mAMS mass concentration is 30% (2σ).⁵⁴ This uncertainty combines the errors associated with particle sampling, the IE measurement technique, assumption of CE, and relative ionization ratio (RIE) of each measured species (20–35% assumed error for total mass).⁵⁴ The total AMS mass concentration correlates well with other measurements (Figure S1). Despite lower resolution than traditional high-resolution (HR) AMS instruments, key ions from the mAMS can be separated at specific mass-to-charge ratios⁵⁵ with a mass resolving power (M/ΔM) of ∼600.⁴⁴

Analysis software, SQUIRREL v1.63H, and PIKA 1.23H processed the mAMS data in the Igor64 7.08 environment (WaveMetrics, Lake Oswego, OR). We modified the mass spectral fragmentation table to separate species based on zero filters’ data.⁵⁶ The modifications were at m/z 15, 29, and 44 to ensure that filter periods show near-zero loading.⁵⁵ All regressions shown in the paper are fit using orthogonal distance regression.

Organic aerosol mass spectral matrices were analyzed using PMF and PMF Evaluation Tool (PET version 3.04a) to differentiate sources.⁵⁷ Analysis of ambient measurements can constrain which primary sources and secondary processes contribute to observed total pollution. The application of PMF to our OA measurements identified unique source factors (mass spectra), which add linearly to reproduce the observed total ambient OA mass spectra and concentration with respect to time. The factors identified by PMF can be associated with sources by comparison to mass spectra of known emission sources or SOA. Thus, the assignment of these factors informs the sources, chemistry, and post-emission aging of ambient OA. Since OA was the bulk of the total measured AMS NR-PM₁ aerosol, this helps determine the origin of most PM₁. Our AMS-PMF calculations on OA use inorganic AMS species and gas-phase measurements for independent comparison to factor times series as external tracers to inform the solution.⁵⁷

Organic mass spectral matrices from all Kathmandu area sites in NAMaSTE2 were combined into one dataset so that PMF factors are consistent across the measurement sites and comparison was straightforward. The combination of sequential multisite OA data into one set for PMF factorization simplified the analysis and significantly reduces the complexity and uncertainty of the solution. The PMF metrics of residuals, mass spectra, and comparison to external tracers determine the optimal solution. The selected solution results from comparisons with robust supporting evidence.⁵⁷ This evidence includes known emission profiles, external tracer time series, and meteorological patterns. In this study, PMF was evaluated in the space of 2–10 factors with FPEAKs ranging from −2 to +2. FPEAKs outside of the range of −1 to 1 failed to converge on a solution. We assume a 5% model error. The FPEAK parameter rotates the solution of PMF through a linear transformation of time series and mass spectra for the different factor size spaces.⁵⁷

Colocated Measurements

At each of the NAMaSTE2 sites, a suite of colocated instrumentation including gas and particle-phase measurements was available. These additional measurements of meteorological and air pollutant concentrations help interpret source apportionment and spatial changes of PM₁.

Several instruments measured gas-phase species. A cavity ring-down spectrometer G2401 (G2401, Picarro, Inc.) measured gas-phase methane (CH₄), carbon dioxide (CO₂), water (H₂O), and carbon monoxide (CO) mixing ratios. Ozone (O₃), nitrogen oxides (NO_x), and sulfur dioxide (SO₂) were measured using Thermo Fisher Scientific, Inc. instruments models 49i, 42i, and 43i, respectively. We performed zero air calibrations on January 5th at Dhulikhel and January 18th at Ratnapark for the NO_x and SO₂ instruments. Meteorological stations (WS700-UMB, Lufft, Germany) recorded ambient temperature, relative humidity (RH), wind direction, and wind speed at 1-min intervals at each measurement site.

Colocated environmental dust monitors (EDM180, GRIMM Aerosol Technik, Germany) ran by ICIMOD at each site provided 1-min optical measurements of PM₁, PM_2.5, and PM₁₀ mass concentrations. This instrument uses a heated inlet to remove water from ambient particles, followed by light scattering at 660 nm to calculate concentrations. This drying mechanism likely also volatilizes some aerosol to the gas phase, artificially lowering concentrations. PM_2.5 filters coincident with the mAMS were collected twice a day for 11-h sample times by the University of Iowa.⁵⁸ Analysis of the filters following collection provided measurements of elemental carbon (EC) and organic carbon (OC) and numerous tracer species used for organic molecular marker-based chemical mass balance (OMM-CMB) source apportionment as described in Islam et al.³¹ A dual-spot aethalometer (AE33, Magee Scientific) measured light-absorbing carbonaceous species in PM. This instrument measures light attenuation by particles on Teflon filter tapes to calculate BC concentration and estimate brown carbon (BrC) absorption.⁵⁹ The AE33 measures eight individual wavelengths (370, 470, 525, 590, 660, 880, 950 nm) and uses filter-loading corrections based on the dual-spot measurement⁶⁰ to calculate real-time concentrations. Aerosol absorption at 880 nm (Babs, 880) and 370 nm (Babs, 370) was used to calculate equivalent BC mass concentrations, using default mass absorption cross sections (MAC) of 7.77 m² g^–1 at 880 nm and 18.47 m² g^–1 at 370 nm.⁶⁰ Absorption due to BC at 370 nm is 2.37 times the absorption at 880 nm. Therefore, light absorption due to BrC is total absorption at 370 nm minus 2.37 times the absorption at 880 nm.³⁵ As a convenient proxy for BrC, we report “BC-equivalent BrC mass” by assuming that the MAC for BrC is the same as the MAC for BC at 370 nm. The AE33 sampled at a 1-s interval averaged to a 1-min interval. The BC measured is assumed to be predominantly submicron based on biomass burning and fossil fuel emissions morphology.⁶¹⁻⁶³ A comparison of the different measurement methods is available in the Supporting Information (Figure S1). The total measured PM₁ aerosol (C-PM₁) is the combined mass of NR-PM₁ from the mAMS plus BC from the aethalometer.

Inlet Setup

A 3-m long 0.6 cm inner diameter copper inlet with a sharp cut PM_2.5 cyclone (Mesa Laboratories, Inc.) sampled at 16.67 LPM, 5 m above ground level for the Dhulikhel and Ratnapark sites, and approximately 25 m above ground level at Lalitpur. The mAMS sampled exclusively from these copper lines. Water was removed from the inlet by a 61 cm Nafion dryer (MD-070, Perma Pure LLC.).⁶⁴ In all cases, the inlet line and laboratory were electrically grounded with a 1-m-long steel rod. A separate system located within 15 m of the mAMS inlet collected PM_2.5 and PM₁₀ filters as discussed in Islam et al.⁵⁸ Gas-phase instrumentation sampled from a dedicated Teflon inlet.

Results Meteorology

The topography in the Kathmandu valley strongly influenced the meteorology. Saddle passes at Chandragiri in the west, and the Karyabinayak river valley in the south are the main avenues of airflow into and out of the valley.⁶⁵ These passes and the valley’s steep walls dictate the flow regime in the valley. The average sunrise and sunset for the NAMaSTE2 campaign were 06:55 Nepal standard (local) time (NST; UTC +05:45) and 17:35 NST,⁶⁶ respectively. The Supplemental Information provides detailed meteorological data (Figures S2 and S3 and Table S3). A semistagnant overnight wind, typically around 0.5 m s^–1 from the east, and temperature inversions from radiative cooling of the surface decrease ventilation-based transport out of the valley.⁶⁵ Thus, overnight emissions are concentrated in a shallow mixing layer about 100 m deep. After sunrise, the mixed-layer depth increases to over 800 m, decreasing concentrations of PM₁ and other pollutants. The primary daytime wind vector was from the west-southwest, and it typically peaked near 3 pm with an average velocity of ∼2 m s^–1.

The temperature at Dhulikhel ranged from a minimum of 2.4 to a maximum of 16.2 °C, with a mean of 7.8 °C. The pressure averaged 838 ± 4 hPa. Relative humidity (RH) had a consistent daily pattern with an average of 67%, a minimum of 33%, and a maximum of 89%. The 24-h average wind speed measured at the Dhulikhel site was 1.3 m s^–1. A morning upslope (anabatic) flow from the north averaged about 1.5 m s^–1 in the early morning and reached 2 m s^–1 at 10:00 NST. The midday wind was light. An evening flow from the north driven by cold regional air settling along the Himalayas averaged 1.75 m s^–1.

Ratnapark had an average temperature of 10.1 °C, ranging from 3.5 to 20.9 °C. The RH at this site averaged 78% (range 30–100%). Multistory apartment buildings and shopping centers border Kanti Path, the major roadway into the downtown area. This urban canyon constrains the immediate wind direction in the Ratnapark area reducing the surface impact of the larger-scale daytime wind vector. The average midday wind speed peaked at 2 m s^–1 at 15:00 NST. Overnight and early morning, the wind speed was relatively constant at 0.5 m s^–1 from the south.

The Lalitpur site’s 24-h average temperature was 11.2 °C (range 4.4–23.5 °C). Winds at the Lalitpur site peak at 2.5 m s^–1 around 15:00 NST from the west-southwest. Overnight the wind slows to an average of 0.25 m s^–1, primarily from the east, with small gusts from all directions.

Pollutant Concentrations: Temporal and Spatial Patterns

Aerosols’ composition and mass concentration varied diurnally and spatially at each measurement site. The average C-PM₁ mass concentrations for each major aerosol species for each site are in Table 1. Gas-phase mixing ratios are in the Supplemental Information (Table S4). The time series and average mass fractions are in Figure 2. On average, C-PM₁ was composed of 49.8% OA, 20.7% BC, and 29.5% inorganic aerosol across all sites. Different aerosol measurement methods from NAMaSTE2 show good agreement (Figure S1). C-PM₁ mass concentrations were ∼1.5 times optical PM₁ (R² = 0.73). Filter-based PM_2.5 was also about 1.6 times optical PM_2.5 (R² = 0.51). The average ratio of optical PM₁ to optical PM_2.5 is 0.85 (R² = 0.73). The C-PM₁ to PM_2.5 filter measurement has a ratio of 0.82 (R² = 0.63), showing good agreement between the different measurement methods. Filter-based PM_2.5 exceeds C-PM₁ by an average of 7.8 μg m^–3 (16%) at Dhulikhel, 21.2 μg m^–3 (21%) at Ratnapark, and 28.5 μg m^–3 (25%) at Lalitpur. These comparisons reflect different size and dust sensitivities, but the agreement is encouraging as detailed in the Supporting Information. The majority of PM_2.5 is PM₁, emphasizing the need for understanding the sources and processes of the submicron size category.

Table 1. Summary of PM₁ Species Mean, 25th and 75th Percentiles, and Percent of Total C-PM₁ from NAMaSTE2 in Nepal 2018.

	Dhulikhel				Ratnapark				Lalitpur
	mean (μg m–3)	25%ile	75%ile	% of total	mean (μg m–3)	25%ile	75%ile	% of total	mean (μg m–3)	25%ile	75%ile	% of total
OA	25.7	21.5	34.4	53.6	45.1	32.2	62.9	44.2	47.9	29.8	50.0	45.2
NH4	3.0	2.6	5.1	6.2	6.8	4.8	7.6	6.7	8.5	6.4	9.6	8.0
Chl	0.8	0.4	1.3	1.6	6.7	2.1	8.0	6.6	4.0	1.0	5.2	3.8
NO3	4.9	3.2	6.5	10.2	7.2	4.5	7.2	7.1	9.2	7.1	10.0	8.7
SO4	4.0	2.3	4.7	8.4	8.2	4.5	9.2	8.0	13.6	10.4	15.6	12.8
BC	7.1	4.2	7.7	14.8	26.4	13.0	31.3	25.9	22.8	12.9	26.8	21.5
BrCa	2.3	1.4	2.7	4.9	1.4	0.5	4.1	1.4	NAb
NR-PM1	40.7	31.0	51.7		75.4	51.7	95.8		84.6	56.9	88.3
C-PM1c	47.8	35.6	58.9		102	68.0	127		106	69.2	114

^aBrC assumed to be a subfraction of OA.

^bBrC data not available at the Lalitpur site.

^cEstimate of total PM₁ (BC + NR-PM₁).

Figure 2.

Time series shown are measured sequentially at three sites: (a) carbonaceous species time series [org, BC, BrC, C-PM₁], inorganic PM₁ [(b) SO₄, (c) NO_3, (d) NH₄, (e) Chl], (f) wind speed and (h) direction, (g) temperature, and (i) RH are from January 4 to February 9, 2018. Wintertime PM₁ average mass fractions are for each measurement site: (j) Dhulikhel, (k) Ratnapark, and (l) Lalitpur.

There was a low boundary layer in the valley early in the morning with nearly stagnant wind speed; this caused a dramatic rise in PM from boundary layer confinement and accumulation after sunrise.⁶⁵ Concentration increases with anthropogenic activity from early morning until the daily afternoon westerly wind forces a minimum. This midday minimum was more pronounced in the city interior than in the suburbs. In the evening, wind stagnation in the valley limited the dilution and transport of particles. PM₁ concentrations increased after the windy period until well after sunset when activity diminishes. Transport from the daily wind, the low morning boundary layer, activity-based emission, and secondary formation from photochemistry drive the daily patterns inside the Kathmandu valley.⁶⁷

Dhulikhel, considered a regional background site with minimal industrial sources, had lower mass concentrations of all species except BrC. The elevated BrC here reflects the increase in biomass burning (BB) in more rural locales. For instance, C-PM₁ at this site averaged 47.9 μg m_,^–3 about half that of Ratnapark or Lalitpur; however, this is still considered an unhealthy level of PM_2.5. Compared to another regional background site measured in previous work, the concentration is higher than the 35 μg m^–3 of PM_2.5 observed at the more remote Godavari site in January 2006.²⁷ The percentage of C-PM₁ for each species at Dhulikhel was like the source regions except for lower chloride and higher BrC. This change in Chl suggests that BB and trash burning (TB) play a more and less prominent role, respectively, at this site but is consistent with the transport of most of the pollutants from the urban source region. Ratnapark, at one of the busiest traffic junctions of the city, has a high density of emission sources, and an average C-PM₁ was 102 μg m^–3, with 45.1 μg m^–3 OA, an estimated 1.4 μg m^–3 of BrC, and 26.4 μg m^–3 of BC. The suburban site, Lalitpur, is the most directly impacted by industry. The measured characteristics were very similar to downtown; C-PM₁ averaged 106 μg m^–3 with a mean OA of 47.9 and 22.8 μg m^–3 of BC. Estimates of BrC concentration were not available for Lalitpur. Diurnal average concentrations are in the Supporting information (Dhulikhel, Table S5; Ratnapark, Table S6; Lalitpur, Table S7).

All three sites exhibited an average daily minimum C-PM₁ concentration, a size category more detrimental to human health, that exceeded the WHO recommended PM_2.5 daily standard of 15 μg m^–3. A combination of activity and topography-mediated meteorology dictates the daily patterns of each site. At Dhulikhel, the background site on the outside edge of the valley, anabatic and katabatic flow drive changes in daily patterns. BC and Chl exhibited the most significant daily variation. BC increased in mass fraction after sunrise and sunset, at traditional cooking times. Chl increases in mass and fraction of total after sunset and until midnight. This pattern is different than the valley interior, where Chl is elevated at night but increases dramatically after dawn. The other PM₁ inorganic species accounted for a roughly constant fraction of the mass.

At Lalitpur and Ratnapark, inside the central valley, prominent morning and evening peaks in daily C-PM₁ concentration occur, which were not observed at Dhulikhel. Urban Ratnapark and suburban Lalitpur exhibit a similar overall diurnal pattern of C-PM₁ (Figure 3). There is a stark contrast between the high concentration in the urban valley and the background at Dhulikhel. The C-PM₁ concentration inside the valley was consistently elevated compared to the background region. There is a greater than 2-fold increase of C-PM₁ from outside to inside the valley. This ratio is the same increase as published in previous work.⁸

Figure 3.

Diurnal profile of C-PM₁ composition (Org, NH₄, NO₃, SO₄, Chl, BC) at (a) Dhulikhel, (b) Ratna Park, and (c) Lalitpur. C-PM₁ mass fraction diurnals for (d) Dhulikhel, (e) Ratna Park, and (f) Lalitpur. Gas-phase mixing ratios (NO_x, NO₂, NO, CO₂, CO, O₃, SO₂) are for (g) Dhulikhel and (h) Ratna Park.

OA was the major C-PM₁ component at all sites across the entire campaign. This large OA fraction is consistent with previous work in various global regions.^68,69 OA at Dhulikhel exhibits minimal variation in daily concentration. However, OA concentrations vary significantly over the day inside the valley, with a minimum during the late afternoon (14:00–16:00 NST). OA concentrations increase both after sunset (17:34 NST) until late at night and after sunrise (06:55 NST) when there is a sudden increase.⁶⁶

Previous measurements in 2015 found a background mixing ratio for CO of 240 ppb at the Bode site and 150 ppb at Dhulikhel,⁶⁷ and a regional background mixing ratio for CO₂ of 411 ppm.^67,70 The CO₂ background at Dhulikhel during this campaign was similar at 408 ppm, but the Ratnapark average mixing ratio was much higher at 450 ppm. The Supporting Information provides gas-phase measurements for NAMaSTE2 (Table S4 and Figure S4).

Gas-phase nitrogen oxides at Dhulikhel averaged 8.9 ppb and ranged from 6.0 to 13.6 ppb (Figure 3g). NO_x at Dhulikhel shows little influence from automobile rush hours, with minimal peaks during these times. O₃, measured from January 4th to 13th at Dhulikhel, had an average 8-h maximum mixing ratio of 53.6 ppb; the daily range for the 8-h maximum was 39.8–57.1 ppb. Dhulikhel had the highest O₃ of the three sites, likely reflecting titration of O₃ by NO_x in the source regions and O₃ production during the transport of pollutants from the source region to the rural surroundings.⁶⁷ O₃ at Dhulikhel showed slight variation, with a consistently elevated mixing ratio. SO₂ showed little variability, averaging 1.1 ppb with a daily low of 0.4 ppb at 06:00 NST and a high of 3.0 ppb at 20:00 NST. CO at Dhulikhel had an average mixing ratio of 601 ppb, with a 10–90 percentile range of 372–783 ppb. The ratio of C-PM₁ to CO (PM/CO) was an average of 0.07 μg m^–3 ppb^–1 with a 10–90 percentile ranging from 0.04 to 0.09 μg m^–3 ppb^–1.

Ratnapark had the highest gas-phase NO_x mixing ratios with an average of 78.52 ppb ranging from 26.2 to 225.2 ppb (Figure 3h). NO_x was unexpectedly high in the morning before sunrise when photochemistry is negligible, possibly from increased heavy vehicle activity in the city overnight. After sunset, there was a second daily peak in the evening, which reached its maximum concentration at 21:00 NST and decreased until 04:00 NST. Evening and morning periods of elevated concentration correspond with rush hour impacts. The average 8-h maximum of O₃ at Ratnapark was only 37.3 ppb, measured from January 14–17 and February 7–29. At Ratnapark, diurnal O₃ was anticorrelated with NO_x. CO measurements at Ratnapark were limited to noon on January 18th to noon on the 20th but averaged 1587 ppb with a 10–90 percentile ranging from 855 to 2555 ppb. CO rose daily from evening till dawn, then slowly decreased over the daytime hours. The PM/CO ratio here was an average of 0.08 μg m^–3 ppb^–1 and had a 10–90 percentile ranging from 0.04 to 0.12 μg m^–3 ppb^–1. Despite limited measurement periods of CO, the PM/CO relationship demonstrates two daily peaks, one concurrent with the overnight rise in CO and one anticorrelated. PM/CO rose daily after sunrise and increased until the wind in the afternoon began. After the wind in the afternoon diminished, PM/CO again rose during the evening, then dropped quickly after sunset. While CO increased all night, PM/CO drops after sunset.

O₃ was more abundant outside the valley than within during this period of measurement. Photochemistry and precursor species’ availability determine O₃ formation.²¹ Based on 2012 winter measurements during SusKat, Sarkar et al.⁷⁰ estimated that NO_x and volatile organic compounds (VOC) from biogenic emissions (24.2%), solvent evaporation (20.2%), traffic (15.0%), and industrial emissions (14.3%) drive ozone production downwind of the valley sources. Overnight the mixing ratio of O₃ within the valley (∼8 ppb) is well below the value outside the valley (∼49 ppb). Elevated urban emissions of NO_x depleted O₃ through titration by reaction with NO. At Ratnapark, there was an increase of the mixing ratio of O₃ in the afternoon and early AM, but not as high as in Dhulikhel. Measurements from 2015 exhibit the same diurnal ozone pattern and mixing ratio in the valley and out.⁶⁷

Organic Aerosol Source Factors

OA makes up an average of 49% of all C-PM₁; therefore, PMF conducted on OA links approximately half of C-PM₁ to its specific sources or processes and provides additional insight into the sources of BC and observed inorganic species. PMF was run on the combined OA time series for all three sites, additional PMF solution selection criteria and diagnostics are in the Supporting Information (Figure S5). Comparing PMF factors to known m/z emissions signatures, correlating with tracer species, and time-based evidence identified those factors and associated sources.^57,71 The seven OA components’ mass spectral profiles for NAMaSTE2 are in Figure 4. All profiles show strong similarities to factors reported in previous studies (e.g., refs (20, 72), and (73)). Uncertainty of PMF solutions is challenging to ascertain, as addressed in the literature.^57,74 A comparison between multiple FPEAK solutions can indicate uncertainty; however, this solution space did not resolve enough FPEAKs to determine uncertainty.^75,76 Absent in this solution is the presence of a primary cooking organic aerosol often seen in US and European datasets.^77,78 The oxidized fragment of m/z 55 (C₃H₃O⁺) is traditionally associated with COA. In regions with elevated HOA, m/z 55 is dominated by C₄H₇⁺, thus COA can be challenging to distinguish.^79,80 Though cooking emissions are certainly present in the valley, the abundance and strength of other sources drive the PMF solution. Previous studies established the difficulty in separating PMF factors with similar time-based usage.⁸¹

Figure 4.

Mass spectra of PMF OA factors for NAMaSTE2, in Kathmandu, Nepal, Winter 2018.

The AMS-PMF OA solution determined the seven factors found in this study (Figure 4), four primary, and three secondary. The four primary factors are hydrocarbon-like OA (HOA), biomass burning OA (BBOA), trash burning OA (TBOA), and a local sulfate-containing OA (sLOA). The secondary factors are oxygenated OA (OOA) 1 to 3 (OOA1, OOA2, OOA3). Diurnal average concentrations are provided in the Supplemental Information (Dhulikhel, Table S8; Ratnapark Table S9; Lalitpur Table S10). Similar factors were identified in NAMaSTE1 at Bode in 2015.¹⁸

In Figure 5, we compare the four primary NAMaSTE2 OA factor mass spectra to the NAMaSTE1 primary sources emission mass spectra. We converted the high-resolution AMS OA PMF factor mass spectra from NAMaSTE1 to unit mass resolution (UMR) to better match NAMaSTE2 resolution to further corroborate and identify factors. The mass spectra measured for primary emission sources in NAMaSTE1 can be seen to closely resemble the NAMaSTE2 AMS-PMF ambient factors.²⁹ A more detailed comparison of the factors comparing relative intensities for all peaks is in Figure S6. The NAMaSTE2 HOA mass spectrum corresponds well to both fresh emissions from idling motorcycles (R² = 0.95) and the ambient high-resolution (HR) AMS-PMF HOA factor (R² = 0.90) measured in NAMaSTE1. This strong correlation is likely due to the abundance of lubricating oil in all engine types. The NAMaSTE1 campaign did not adequately characterize all vehicle emissions; however, these HOA factors share characteristics with other extensive urban studies (e.g., ref (73)). The NAMaSTE2 HOA ratio of m/z 55:57 (C₄H₇⁺ to C₄H₉⁺) is 1.03. A typical HOA m/z 55:57 is just below 1:1,⁶⁴ such as from NAMaSTE1, where it was 0.95.²⁰ While this ratio is elevated, the NAMaSTE2 HOA times series (Figure 6) correlates well with other primary vehicle emissions such as NO_x (slope = 0.09 μg m^–3 ppm^–1, R² = 0.65) and CO (slope = 0.01 μg m^–3 ppb^–1, R² = 0.75) (Figure S7).

Figure 5.

Mass spectra from NAMaSTE1 source testing, and NAMaSTE1&2 ambient AMS-PMF primary OA factors: (a) HOA, HR HOA, and motorcycles; (b) BBOA, HR BBOA, and hardwood and mixed agricultural burning; (c) TBOA, HR TBOA, garbage, chip bags, and mixed plastic burns; and (d) sLOA, HR sLOA, clamp, and zigzag brick kilns, and coal cookstoves.

Figure 6.

Source factor OA mass concentration and collocated time series of (a) HOA and black carbon; (b) BBOA; (c) TBOA; (d) sLOA and sulfate; (e) OOA1, 2, and 3; (f) gas-phase CO; (g) brown carbon; (h) particulate chloride; gas-phase (i) SO₂, (j) NO_x and (k) O₃ from NAMaSTE2 in Nepal for January 4 to February 9, 2018.

The NAMaSTE2 BBOA factor has similarities to mixed agricultural burning emissions from NAMaSTE1 (R² = 0.95). The BBOA mass spectrum also correlates with the ambient HR AMS-PMF BBOA factor from NAMaSTE1 (R² = 0.73). The BBOA concentration times series correlates well with biomass burning associated species: BrC (slope = 0.2, R² = 0.62), CO (slope = 0.0034 μg m^–3 ppb^–1, R² = 0.63), CO₂ (slope = 0.115 μg m^–3 ppm^–1, R² = 0.64) and AMS levoglucosan measurements (slope = 14.424, R² = 0.95). The biomass burning marker ion m/z 60 (C₂H₄O₂⁺), which comes from burning levoglucosan or other anhydrous sugars, is a significant indicator for BBOA.⁸² The NAMaSTE2 BBOA ratio of m/z 55:57 was 1.16, lower than expected compared to NAMaSTE1 (1.79), but still elevated above 1:1.

TBOA from NAMaSTE2 shows similarities to the HR AMS-PMF factor TBOA from NAMaSTE1 (R² = 0.63) and both mixed garbage (R² = 0.68) and plastic burning (R² = 0.72) from source measurements in NAMaSTE1. The NAMaSTE1 and 2 TBOA factors correlate well. NAMaSTE2 TBOA correlates with the tracers BrC (slope = 0.152, R² = 0.52) and CO (slope = 0.0035 μg m^–3 ppm^–1, R² = 0.52).

The sLOA factor from NAMaSTE2 correlates best with the 2015 NAMaSTE1 HR AMS-PMF sLOA at Bode (R² = 0.81), with lower agreement with the direct primary source mass spectra from clamp style (R² = 0.35) and zigzag style brick kilns (R² = 0.35) and charcoal (R² = 0.57) source emissions. The NAMaSTE1 sLOA factor had better correlations with these emission spectra than NAMaSTE2 sLOA (charcoal cooking (R² = 0.70), clamp style brick kilns (R² = 0.43), and zigzag style brick kilns (R² = 0.35)¹⁸). The sLOA time series correlates with m/z 85, a coal tracer (slope = 13.325, R² = 0.69),³⁵ SO₄ (slope = 0.473, R² = 0.57), and SO₂ (slope = 0.152 μg m^–3 ppm^–1, R² = 0.45) and the highest sulfur emissions were observed from brick kilns in NAMaSTE1.^41,42 The correlation of sLOA with coal and brick kilns is suggestive that sLOA is from coal-burning brick kilns.

In general, the inorganic mAMS species data support our assignment of OA source factors. SO₄ is the form of sulfur most prevalent in the aerosol phase.⁸³ Primary emissions from coal combustion and brick kilns are the suspected dominant source of aerosol sulfate in the Kathmandu Valley; the burning of agricultural residue contributed to a small degree.^29,42 SO₄ as measured by the AMS has low correlation with BBOA (slope = 0.54, R² = 0.17, Figure S8), but correlates well with sLOA (slope = 0.47, R² = 0.57). Brick kilns have a large sulfate emission factor.⁴¹ Nitrate is typically formed from the oxidation of atmospheric NO₂ and has no major primary sources. However, high nitrogen content in coal can lead to elevated NO_x.⁴¹⁴¹ NO₃ here trends well with sLOA (slope = 0.57, R² = 0.64). This correlation may be due to high nitrogen content in coal or an artifact due to sLOA following the same general temporal trends as C-PM₁. The sum of OOAs correlates well with NO₃ (slope = 1.67, R² = 0.61), although individual OOAs have a low correlation to NO₃. Ammonia is typically associated with animal waste, soil emissions, fertilizers, and industrial emissions; however, in the Kathmandu Valley, cookstoves, coal, brick kilns, and agricultural burning are the major sources of submicron aerosol ammonium.^35,41,42,83 NH₄ has time-based similarities to HOA (slope = 1.62, R² = 0.23), BBOA (slope = 0.79, R² = 0.33), and sLOA (slope = 0.47, R² = 0.57). In the Kathmandu Valley, dung, and charcoal-fired cookstoves, brick kilns, coal combustion, and the open burning of garbage and agricultural residues are the major sources of atmospheric chloride.^35,39,41,42 Chl is however poorly correlated with TBOA (slope = 0.45, R² = 0.23), HOA (slope = 1.58, R² = 0.36), sLOA (slope = 0.59, R² = 0.34) and BBOA (slope = 0.78, R² = 0.47). This suggests that Chl is mostly from BB, or that much of the Chl was lost to due dilution or aging.⁸⁴⁻⁸⁶

The NAMaSTE2 OOA factors compare well to OOA factors from NAMaSTE1 (Figure S9). The OOA1 mass spectra have strong similarities with a slope of 1.45 and R² of 0.94. OOA2 differs significantly (slope = 0.37, R² = 0.30), which may be due to differences in season and potential influences from aging of the additional biomass combustion in the winter compared to the pre-monsoon period. OOA3 from NAMaSTE1 and NAMaSTE2 show some similarities with an R² of 0.50. The total NAMaSTE2 OOA concentration correlates well with m/z 44 with respect to time (slope = 3.88, R² = 0.83).

Organic Aerosol Temporal Trends

Traffic and other fossil fuel uses have a more considerable impact on the dense urban environment than the background regional site. The average and 10th/90th percentile for OA factors for each site is in Table 2.

Table 2. Organic Aerosol PMF Factor Summary from NAMaSTE2.

	Dhulikhel				Ratnapark				Lalitpur
	mean (μg m–3)	10% (μg m–3)	90% (μg m–3)	%OA	mean (μg m–3)	10% (μg m–3)	90% (μg m–3)	%OA	mean (μg m–3)	10% (μg m–3)	90% (μg m–3)	%OA
HOA	2.3	0.5	3.0	9.00	11	1.6	17	23.7	11	1.1	21	23.9
BBOA	2.3	0.6	3.1	9.10	8.3	1.6	12	18.3	6.2	1.4	9.7	13.0
TBOA	3.8	1.3	4.2	14.7	6.3	2.4	7.2	14.1	3.1	0.9	3.7	6.40
sLOA	2.7	0.7	3.2	10.4	4.6	1.2	6.9	10.3	7.9	2.2	10	16.5
OOA1	8.5	3.4	8.3	32.9	6.3	2.7	6.2	14.0	10	4.0	10	21.7
OOA2	2.7	1.0	3.2	10.5	3.8	0.6	5.8	8.40	5.7	0.4	10	11.9
OOA3	3.4	1.3	3.9	13.4	5.1	1.8	6.3	11.3	3.2	0.9	4.1	6.70

Multiple events of elevated HOA at Ratnapark occurred with concentrations over 60 μg m^–3. HOA averaged 10.7 μg m^–3 and made up 24% of the OA mass inside the valley at Ratnapark. BBOA averaged 8.3 μg m^–3 and made up 18% at Ratnapark. Isolated BBOA plumes inside the city often reached concentrations over 30 μg m^–3. Enhancement in the BBOA fraction of total OA across all sites is associated with localized biomass burning. Ratnapark had numerous vendors with wood-fired food stalls near the measurement site. This source was unique to that site as food vendors were far less prevalent outside the urban center. TBOA made up 14% of OA at Ratnapark with an average concentration of 6.3 μg m^–3. Ratnapark sees an average maximum daily TBOA concentration over 7.0 μg m^–3, while the other two sites have an average maximum of 4.0 μg m^–3. The sLOA components contributed 10% of OA at Ratnapark with an average concentration of 4.6 μg m^–3. The combined OOA factors made up 33.7% at Ratnapark. Averaging 6.3 μg m^–3, OOA1 was the most significant OOA component at Ratnapark, 43% of OOA. OOA2 accounted for 24% of OOA, an average of 5.7 μg m^–3 at Ratnapark. OOA3 averaged 3.2 μg m^–3 and made up 33% of OOA at Ratnapark. OA at Ratnapark, within the city center, follows the same diurnal peak pattern as C-PM₁. There is a significant increase immediately after sunrise and a second peak in the evening. This morning peak dramatically increases HOA, BBOA, and OOA2 concentrations. Ratnapark has the highest fraction of primary OA; HOA is the largest single fraction. SOA is only the majority of OA during the period of midday with elevated wind speed. The HOA concentration and mass fraction increase dramatically during AM and PM rush hours. During these periods, HOA accounts for over 25% of all OA. However, recent work has indicated that default AMS quantification parameters may overestimate this reduced organic aerosol by a factor of 2 for ambient studies with fresh sources.⁸⁷ TBOA has a similar mass fraction and concentration at Ratnapark and Dhulikhel. The mass fraction peaked during the period of lowest overall C-PM₁, and concentration maximized during the morning OA peak. BBOA at Ratnapark had peaks in mass fraction and concentration in the evening around 21:00 NST, which decreased steadily until sunrise. The concentration then increased until the wind began to flow. The peaks in BBOA correspond to morning and evening mealtimes, and many food stands and open grills were in use in the urban area. Secondary organic aerosol concentration in OOA2 increases just after sunrise, followed about an hour later by increasing OOA3. sLOA had a sizeable daily increase in mass fraction and concentration immediately after sunrise and reached a maximum concentration before the directional shift and increase in wind speed.

At the suburban site of Lalitpur, there is a dual peak diurnal pattern in OA concentration, similar to observations at Ratnapark. The morning peak, after sunrise, had a significant increase in OOA2 and BBOA. After the wind speed decreases in the evening, sLOA builds up and reaches a high concentration making up a large mass fraction until sunrise. HOA influences Lalitpur during rush hour. OOA2 increases the most following dawn and remains a large fraction until the late evening when other primary sources increase concentration. This locality has the most negligible impact from TBOA; refuse collection was more common in this region. HOA made up 10% of OA, a mean of 11.4 μg m^–3, and BBOA averaged 6.2 μg m^–3, 13% of OA. HOA made up 24% of OA mass inside the urban areas, compared to 10% in rural regions. Traffic and other fossil fuel consumption had a more considerable impact on the dense urban environment than at the Dhulikhel site.

Notably, while the TBOA contribution to OA was similar at both Dhulikhel and Ratnapark, it was only half that, 6% (3.1 μg m^–3), at Lalitpur. This difference in trash burning contribution may be due to differences between the peri-urban residential Lalitpur site and other sites. Ratnapark and Dhulikhel see a similar background level of approximately 2.0 μg m^–3. However, burning activity around the Ratnapark measurement site leads to a higher maximum daily TBOA concentration of over 7.0 μg m^–3. In contrast, the other two sites have a maximum TBOA concentration of 3 μg m^–3. In another notable discrepancy, the sLOA components make up 10% of OA at Dhulikhel and Ratnapark but 17% at Lalitpur, with an average concentration of 7.9 μg m^–3. During typical daytime wind conditions, Lalitpur is downwind of several nearby brick kiln sites. sLOA reached its maximum for the campaign of 33.8 μg m^–3 at 09:00 NST on February 2nd at Lalitpur when the wind was from the ENE. This wind vector is directly from a major brick kiln region.

The combined oxygenated OA (OOA) factors made up 40.3% of OA at Lalitpur. Similar to Dhulikhel, OOA1 was the main component of OOA, averaging 10.4 μg m^–3, and made up over half of OOA at Lalitpur. OOA2 averaged 5.7 μg m^–3 and accounted for 30% of OOA at Lalitpur. OOA3 averaged 3.2 μg m^–3 and made up 15% of total OOA at Lalitpur. In general, Table 3 shows that as the distance from the primary downtown source region increased, OOA/OA and OOA1/OOA increased as well.

Table 3. Oxygenated Organic Aerosol Wintertime Urban Gradient.

	Ratnapark	Lalitpur	Dhulikhel
distance from City Center (km)	0.5	6.5	32
OOA:OA	0.34	0.40	0.57
OOA1:OOA	0.41	0.53	0.58

As with the time series’ the campaign average diurnal patterns for OA concentration and mass fraction, along with campaign averages at each site (displayed in Figure 7a–i), show variation due to source changes and meteorological influence. The more urban areas experienced more significant variation in daily organic aerosol concentrations than the Dhulikhel site, which had little dynamic change (Figure 7a–c). The mass fraction of each AMS OA factor exhibited a minimum diurnal variation at Dhulikhel and significant change at the Ratnapark and Lalitpur sites, which have more substantial source emissions (Figure 7d–f). The low background of NO_x and VOC observed at Dhulikhel limits the formation of new SOA from photochemistry.⁶⁷ This limitation is consistent with the minimal change in the OOA fraction observed at Dhulikhel. SOA was likely made during transport from the downtown region to Dhulikhel.

Figure 7.

OA source components from PMF for NAMaSTE2 in Nepal, January 4 to February 9, 2018. The average diurnal mass concentration at (a) Dhulikhel, (b) Ratna Park, and (c) Lalitpur and diurnal mass fraction for (d) Dhulikhel, (e) Ratna Park, and (f) Lalitpur, as well as an average mass fraction for (g) Dhulikhel, (h) Ratna Park, and (i) Lalitpur are displayed.

The bulk aerosol at Dhulikhel was the most aged and photochemically processed of all sites. Despite being removed from the brick kiln region, the sLOA factor remains a significant portion of the overall mass, with a similar mass fraction as the city center, indicating this factor’s ubiquity throughout the valley and greater region.

OOA is the largest fraction of total OA mass at all sites, consistent with other urban environments.⁶⁹ The combined oxygenated OA (OOA) factors were ∼57% of OA in the more background rural environment of Dhulikhel and lower fractions in the more primary source heavy regions of Ratnapark (∼34%) and Lalitpur (∼40%). The average mass fraction of primary to secondary OA is highest inside the urban core at 63%, where the known precursor sources are, compared to 43% at Dhulikhel on the valley edge.

OOA3 shows little change in mass concentration across sites and minimal diurnal variation. It makes up 25% of OOA at Dhulikhel, 33% at Ratnapark, and 15% at Lalitpur. OOA2 accounts for 19% of OOA at Dhulikhel, 24% at Ratnapark, and 30% at Lalitpur. OOA2 exhibits increased concentration and mass fraction just after sunrise and a similar concentration at all locations. OOA1 is the main component of OOA and makes up 58% of OOA at Dhulikhel, 43% at Ratnapark, and 55% at Lalitpur. At all sites, OOA1 is the dominant fraction during the afternoon, especially in the valley during the period of the highest wind speed. Outside the valley, OOA1 is always the single most significant factor. OOA1 is likely an aged regional SOA brought into Kathmandu through long-range transport.⁶⁵ This factor shows a substantial change in concentration during the increase in elevation of the boundary layer after sunrise and during peak ozone. Previous work on secondary formation in the Kathmandu valley estimated SOA production at Bode in the winter of 2012 to come from brick kilns (28.9%), traffic (28.2%), industry (25.7%), and residential biofuel (12.8%).⁷⁰ While these are not directly comparable to the source factors determined herein, there are similarities. Brick kilns as sLOA and traffic as HOA are major sources.

Black Carbon Source Estimation

The average BC concentration in densely populated areas of the valley was 2–7 times greater than at the valley edge. The BC diurnal averages at all sites have two distinct peaks in mass fraction. The initial rise in mass fraction and concentration is earlier (07:00 NST) in the valley than outside (09:00 NST), suggesting transport to the perimeter. The second peak is at an average of 20:00 NST for all sites suggesting evening rush hour and biomass burning for home heating and cooking. The concentrations and patterns of these peaks and afternoon minima are consistent with previous measurements made within Kathmandu.^20,88 A restriction forbids trucks from passing through the city limits during daytime after 07:00 NST. In addition to wind-driven dilution, this source control reduces daytime BC concentrations.

BrC is associated with biomass burning. BrC concentration was 1.6 times greater inside the valley than at Dhulikhel. BrC in the urbanized valley was elevated above the background level, with a maximum in the evening at the same time as the peak observed in Dhulikhel.

BC comprised 21% of all C-PM₁, making it the second largest fraction of the total mass after organic aerosol. The combination of OA and BC accounted for 70% of all C-PM₁. Attributing these species to their respective sources thus accounts for an even larger fraction of the submicron particulate matter in the Kathmandu Valley than OA alone. We used measured ratios of OA to OC (Table S11) and OC to BC (OC/BC; Table S12) for specific sources from NAMaSTE1.^35,42 We follow the methodology of Werden et al.²⁰ and use the OA converted to OC along with literature OC/BC ratios for an estimate of the BC contribution from these sources (Figure 8). This estimate is independent of the measured BC concentration. The estimated BC contribution from each OA source factor is summed to the total (BC_est; see eq 1) and compared to the measured BC to see if the method can reconstruct the BC mass. BC correlation to the reconstructed BC_est from OA factors have significant similarities (slope = 0.92, R² = 0.49) (Figure 10).

1

Figure 8.

BC reconstruction (a) individual factor and BC measured time series. Average BC mass fraction for (b) Dhulikhel, (c) Ratna Park, and (d) Lalitpur, Individual factor average and range for (e) Dhulikhel, (f) Ratna Park, and (g) Lalitpur, from NAMaSTE2 in Winter 2018, Kathmandu, Nepal.

Figure 10.

Mass fraction (a), time series (b), and mass spectra (c), for Chl spike on January 24th, at Ratna Park in Kathmandu, Nepal.

Due to the similarities in the OA source factors, the same OC/BC ratios from NAMaSTE1¹⁸ are used. These are available in the Supporting Information (Table S12). Comparing the BC contributions from different sources across sites yields interesting results. In Dhulikhel, the lower-traffic regional site is downwind of Bhaktapur, which has a high brick kiln density, BC associated with sLOA, from brick kilns or coal use, was the main contributor at 49% of the total. The BC associated with traffic-related HOA was the second most significant factor at the background site (41%). Biomass and trash burning contributed 4.3 and 5.5% of the total BC at Dhulikhel. BC associated with HOA from vehicle emissions in higher population and traffic density sites dominates the total BC contribution at Ratnapark (62%) and Lalitpur (55%). BC associated with coal or brick kilns from the sLOA factor has the second most significant impact at these sites, with 29.8% at Ratnapark and 41% at Lalitpur. The biomass and trash burning contributions to the total estimated BC were less than 5% at the urban and suburban sites and lower than the values for the background at Dhulikhel.

This reconstruction shows the impact of each major source on the total BC. The main source of BC in the Kathmandu Valley was traffic. HOA and sLOA sources have a very significant effect on total BC. In related work, diesel vehicles and trucks were major BC sources in India; in the IGP, trucks are less than 20% of the vehicle fleet but account for up to half of vehicular BC emissions.⁸⁹ In Nepal, BC emissions from HOA-related traffic sources are likely also dominated by poorly controlled truck emissions. We found that in the Kathmandu Valley, vehicular emissions made up 52% of BC and 23% of OA. Only 2.3% of registered vehicles in Kathmandu are trucks, but due to very high emissions factors and usage, this small percentage can account for more than 80% of vehicle emissions for PM_2.5.⁹⁰⁹⁰ Enforcing the EURO VI standard for vehicle emissions could have considerable impacts in the valley, with a potential for 30% less CO₂ and a 44% decrease in other emissions.²⁶

Colocated Organic Aerosol Apportionment Intercomparison

NAMaSTE2 conducted two independent source apportionments for OC in the Kathmandu Valley. The AMS-based effort, as described above, used the similarity of mass spectra of PMF factors with known submicron mass spectra of primary and secondary sources to identify and quantify source contributions. The complementary approach used offline molecular marker-based chemical mass balance analysis of PM_2.5 from colocated filter sampling.⁵⁸ As detailed next, both approaches identified the same major contributors. We examine the agreement to assess our overall confidence in sources and ultimately inform mitigation strategies. To directly compare OA to OC, measured ratios from source experiments (OA/OC³⁵) were used to convert OA for each PMF factor to OC (Table S1).

Known speciated emissions profiles underpin the calculations of OMM-CMB OC source contributions. The sources used are vegetative detritus, traffic, biomass, garbage burning, coal burning, and secondary organic carbon (SOC) tracers. Vegetative detritus was a minor component of the filters. A direct comparison of these sources to AMS-PMF OC factors shows good agreement, as shown in Figure 9c–g.

Figure 9.

OC source factors: (a) CMB time series, (b) PMF times series for NAMaSTE2, January 4 to February 9, 2018. Correlations for CMB and PMF OC factors: (c) HOA vs traffic, (d) BBOA vs biomass, (e) TBOA vs trash, (f) sLOA vs coal, and (g) OOA vs SOC.

The regression of HOA OC and filter-based traffic OC factors has a slope of 1.98 (R² = 0.32); this is consistent with the observed increase in relative ionization efficiency for reduced organic components in the AMS instrument.^87,91,92 An increased RIE would explain the discrepancies between AMS-PMF HOA OC and OMM-CMB traffic OC. However, we could not perform HOA IE calibrations with this instrument and have, therefore, not changed the relative ionization efficiency (RIE) value used for HOA in the analysis. As is typically observed, the AMS and filter comparison is closer for factors associated with more oxidized aerosol. For instance, biomass burning OC from AMS-PMF and OMM-CMB correlate well with a slope of 1.27 and R² of 0.66. The trash burning factors are similarly well associated, with a slope of 1.143 and R² of 0.22.

On the other hand, SOC derived from the sum of OOA shows a positive mass bias toward the OMM-CMB factor, with a slope of 0.5 and R² of 0.07. The OMM-CMB calculation of SOC does not capture all SOA because molecular markers exist only for SOA from specific precursors.⁵⁸ Thus, we have added the unassigned OC from filters to measure SOC in this comparison. This improves the comparison between AMS SOC and filter-based SOC. While there are differences between these factors, the sum of HOA, sLOA, and OOA OC compares better to the sum of traffic, coal, SOC, and other (slope = 0.93, R² = 0.34; Figure S11). This increases our confidence in the apportionment of BBOA and TBOA, but the coal OMM-CMB factor and sLOA show the most significant difference, with almost 30 times more mass in the AMS-PMF solution. During NAMaSTE1, at Bode in 2015, the sLOA factor OC vs OMM-CMB coal OC slope was 1.62²⁰ compared to 27.89. This large discrepancy is potentially attributable to a different coal source in 2018 compared to source measurements made in 2015. During the 2015 NAMaSTE1 source measurements, Nepal imported 50% of coal from India, 16% from Indonesia, and 26% from South Africa. In 2018, all imported coal came from India.³³ Using different coal with a reduced concentration of the indicator species used for the OMM-CMB solution could impact the ratio expected. Overall, the two source apportionment approaches identify traffic, biomass burning, and garbage burning as the three main sources of fine particulate carbon in the wintertime and targets for mitigation. The agreement on the latter two sources is better than 30%. The filter-based study also implicates dust as a significant source, as Islam et al.⁵⁸ detailed.

Inorganic Observations

Sulfate was the most prominent anionic inorganic submicron aerosol species within the valley. Brick kilns constitute a significant sulfur source in South Asia.⁹³ SO₄ diurnal patterns at Dhulikhel and Ratnapark showed minimal daily variation. Lalitpur, the site closest to the brick kilns, had elevated SO₄ concentrations and mass fractions compared to the other sites. There is a substantial gradient in SO₄ concentration that increases from Dhulikhel to the urban center. The highest SO₄ concentration measured was 63.5 μg m^–3 at Lalitpur on the 4th of February at 09:06 NST. During this period, the m/z 55:57 ratio was less than 1, and m/z 60 was elevated. SO₄ and Chl spectral patterns were similar to the fresh emissions of clamp brick kilns.

Nitrate was the dominant anionic inorganic species at Dhulikhel. The maximum nitrate concentration across the campaign was 45.5 μg m^–3, observed in Lalitpur on February 2nd. NO₃ near the valley exterior shows a decreasing concentration from 05:00 NST to 18:00 NST, with an increase overnight until the 05:00 NST maximum. NO₃ inside Kathmandu had a distinct concentration peak around noon after the daily AM PM₁ peak. The mass fraction of NO₃ was maximized from noon until the early evening. The increase of nitrate was more pronounced inside the urban center than the suburban site.

Ammonium, the principal inorganic cation in C-PM₁, generally balanced the anionic inorganic species. During the campaign, the maximum concentration reached was 40.8 μg m^–3 at Ratnapark on the night of January 24th. This increase corresponds with an increase in Chl, described below. NH₄ mass fraction diurnal patterns were similar at all three sites and showed minor variability.

Chloride was the most variable between sites, with concentrations at Ratnapark being up to 6 times greater than Dhulikhel. Chl outside the valley shows the lowest average concentrations, with a slight increase in mass fraction overnight. The Chl diurnal pattern at Ratnapark had two prominent mass fraction and concentration peaks, the first at 10:00 NST and the second at 22:00 NST. Chl at Lalitpur is not as elevated as Ratnapark but showed the same diurnal mass fraction trend. The increase in evening Chl was much more pronounced at Ratnapark, potentially driven by the end-of-workday burning. The lowest concentrations were observed between 10:00 NST and 16:00 NST during daylight. Concentrations of Chl increased rapidly overnight to almost 3.0 μg m^–3. Gas particle partitioning is an essential mechanism in particulate chloride concentrations. A smaller chloride fraction is expected in the gas phase during colder periods.⁴⁰ The elevated chloride concentration and lack of neutralization suggest that either external mixtures of chloride-containing particles exist or the potential of organic chlorides. The maximum campaign concentration of Chl of 120 μg m^–3 occurred on the evening of the 24th at Ratnapark. The Chl of this plume was about 20 times more concentrated than background conditions at Ratnapark. These Chl concentrations are among the highest ever measured, similar to those in New Delhi, India.⁴⁰ This plume is potentially linked to substantial poly(vinyl chloride) (PVC) burning, which, unlike combusted plant material, can emit more chlorides than carbonaceous species.⁹⁴

This elevated chloride event was an enhancement of OA followed by an increase of inorganic Chl and NH₄. This 1-min incident increased OA by 85.7 μg m^–3 above the background, to a total of 174 μg m^–3. This OA increase is mainly OOA2 but also includes BBOA, TBOA, and HOA. The organic mass spectra had a significant contribution from m/z 60, about 2% of OA. OOA2 is enhanced by 19.1 μg m^–3, BBOA by 12.3 μg m^–3, TBOA by 9.6 μg m^–3, and HOA by 7.2 μg m^–3. One minute after the OA rise, Chl and NH₄ increased. Chl accounted for 40% of C-PM₁ during this plume. The C-PM₁ concentration of this event was 257 μg m^–3 (Figure 10). There was no change in wind speed or direction during this event, nor did BC and BrC have a corresponding increase. The separation in time from the impact of the OA and inorganic plumes indicates the potential heterogeneity of the burning initiated with biomass and then with plastic or PVC containing refuse. NH₄ reached a maximum concentration during this event of only 40.7 μg m^–3, meaning the majority (64%) of the aerosol was not neutralized.

The aerosol neutralization ratio (ANR) is the ratio of aerosol cation charges to anion charges, assuming that ammonium is the only cation present. Zhang et al.⁹⁵ defined ANR as

2

The ANR at Dhulikhel averaged 0.99, or mostly neutralized, whereas ANR was less than one at Ratnapark, where it averaged 0.83. At Lalitpur, the ANR averaged 0.79, ranging from 0.24 to 1.19. The low ratio of inorganic anion to cation species in the valley means that the bulk of the aerosol mass was not fully neutralized, or an additional cation was present. This lack of full neutralization indicates an increase in the acidity of the aerosol. Previous studies found aerosol pH in Kathmandu ranged from 2.2 to 3.3.^31,68,96

VOCs reacting with nitrate radicals to form organic nitrates (OrgNO₃) were an additional source of secondary OA.⁹⁷ The campaign average ratio of the high-resolution ions NO₂⁺ to NO⁺, shown in Figure S12, was 0.92. This average value is like the average 0.95 (R_calib) ratio of NO₂⁺/NO⁺ for pure NH₄NO₃ from calibrations. A measured ratio of NO₂⁺/NO⁺ below the calibrated value acts as an indicator of OrgNO₃ formation.⁹⁸ The ratio for each site is shown in Table 4. The concentration of organic nitrate is calculated using this ratio (R_measured), assuming a detection limit of 0.1 μg m^–3 and an estimated uncertainty of ±20%.⁹⁷ OrgNO₃ is defined as⁹⁷

3

Table 4. Ratio of NO₂⁺ to NO⁺ and Organic Nitrate Average and 10th and 90th Percentiles for the Three Sites from NAMaSTE2 in Kathmandu, Nepal, 2018.

	Dhulikhel			Ratnapark			Lalitpur
	mean	10%	90%	mean	10%	90%	mean	10%	90%
NO2+/NO+	0.88	0.78	0.98	0.98.	0.83	1.14	0.93	0.81	0.02
OrgNO3(μg m–3)	0.74	0.27	1.26	0.57	0.05	1.18	1.09	0.02	2.13
OrgNO3/NO3+	0.08	0.03	0.14	0.07	0.03	0.13	0.07	0.03	0.11

Organic nitrate accounts for an average of 7% of total nitrate, with a campaign maximum of 38%. The average fraction of OrgNO₃ to NO₃ in the Kathmandu Valley was much lower than measured in European urban environments (39%⁹⁷). The highest concentrations of OrgNO₃ measured during the campaign were at the Lalitpur site, where the maximum concentration reached 5.4 μg m^–3 at noon on February 7th. The daily peak at all three sites is after dawn but before the afternoon wind increase. Concentrations of OrgNO₃ follow the NO₃ pattern at all sites and increase as soon as the wind diminishes. This corroborates the findings of Yu et al.⁹⁹ that OrgNO₃ was generated locally and not transported from outside the region.

Conclusions

Sequential air quality measurements at three sites in the Kathmandu valley, Nepal, in January and February of 2018 showed significantly elevated PM₁ concentrations. Compared to the rural valley rim at Dhulikhel, 2–4 times higher concentrations were later observed within the urban valley center at Ratnapark and Lalitpur. The high density of sources in the more populated regions leads to extreme pollution levels and poor air quality. The largest AMS species, OA, was apportioned to factors by PMF, identifying traffic, biomass burning, trash burning, and coal use in brick kilns as the main contributors. Cooking aerosol related to food-related emissions (e.g cooking oils as observed in other areas) did not have a discernable influence here due to the strength of other emission sources. These factors agreed with filter-based source apportionment of PM_2.5 OC. Conversion of those factors from OA to BC estimated the contribution to BC by each source. This BC apportionment identified traffic, likely heavy trucks, and the coal use in brick kilns as the main contributors to elevated concentrations. The concentration and chemical composition of PM₁ varied by time of day, influenced by photochemistry, meteorology, and patterns of activity. PM₁ concentrations were highest during the morning, post sunrise, and minimized during the late afternoon when ozone mixing ratios and wind speed peaked.

Across all diurnal cycles, OA was the most significant component of PM₁ and BC the second largest. Secondary formation of organic aerosols was abundant, making up approximately 40% of OA averaged across all sites. While the measurements were not made coincident in time, measurements at the background Dhulikhel site showed the high background concentration of the regional aerosol (∼50 μg m^–3). The influence of primary emissions was diminished on the valley edge at Dhulikhel, while the impact of SOA was relatively enhanced. The emission of vehicles heavily impacted Ratnapark. However, the combustion of trash and biomass near this site contributes to overall OA more than vehicles did. Lalitpur, the suburban site, was subject to similar OA loading as the city center but with a much-diminished impact from biomass and trash combustion. The impact of organic sulfur and organic nitrates was highest at this site, as was OA from coal or brick kilns.

During the period of measurement within the valley, PM was generated locally. Vehicles, brick kilns, and trash burning represented 50% of all OA and 95% of BC emissions. BC and OA made up most C-PM₁, about 70% of the total. Though there are limitations due to the serial nature of these measurements, they show that vehicles, coal use, and trash burning are ready targets for emission controls. Reducing coal combustion in brick kilns and cookstoves could reduce OA within the valley by over 10% and BC by 30%. Reducing brick kiln emissions would also control organic sulfur and sulfur dioxide. Trash burning is a regional source of OA and chloride species, this evidence of open burning suggests that trash control or collection must be better addressed. Vehicles are responsible for up to 24% of OA and potentially 60% of BC and could be more adequately controlled with modern fuel standards. The diminished traffic signature at Dhulikhel and elevated urban concentrations suggest that mitigating these sources in the valley will likely improve air quality to a healthier level.^100,101

The Kathmandu valley contains a gradient of urban density and source concentrations, from downtown and the international airport to rural farmland. This variation changes the population’s exposure to local air pollutants. Further research measuring multiple locations without the temporal differences from sequential measurements would significantly increase knowledge of south Asian air quality issues. Our coordinated measurement of multiple gas-phase species and PM speciation across multiple consecutive sites creates a more robust understanding of the valley’s wintertime air pollution sources, identifying brick kilns, trash burning, and traffic as ready sources for stronger emission controls.

Glossary

Abbreviations

ANR

aerosol neutralization ratio

Babs λ

absorption coefficient for wavelength λ

BBOA

biomass burning organic aerosol

BC_est

estimated black carbon

BrC

brown carbon

CE

collection efficiency

CH₄

methane

Chl

aerosol chlorides

CO

carbon monoxide

CO₂

carbon dioxide

C-PM₁

total submicron aerosol

DoEnv

Nepali Department of Environment

d_va

vacuum aerodynamic diameter

EC

elemental carbon

ePToF

enhanced particle time of flight

H₂O

oxidane

HEPA

high-efficiency particle air [filter]

HOA

hydrocarbon-like organic aerosol

ICIMOD

International Center for Integrated Mountain Development

IE

ionization efficiency

IE/AB

ionization efficiency/airbeam

IGP

Indo-Gangetic Plain

MAC

mass absorption cross section

MS

mass spectra

NAMaSTE

Nepal ambient monitoring and source testing experiment

NO_x

nitrogen oxides

NR-PM₁

nonrefractory submicron aerosol

NST

Nepali standard time

O₃

ozone

OA

organic aerosol

BC

black carbon

OA/OC

organic aerosol-to-organic carbon ratio

OC

organic carbon

OC/BC

organic carbon-to-black carbon ratio

OMM-CMB

organic molecular marker-based chemical mass balance

OOA

oxygenated organic aerosol

OrgNO₃

organic nitrate

PM

particulate matter

PMF

positive matrix factorization

PToF

particle time of flight

Q

PMF scaled residuals

Q_exp

PMF expected residuals

R_calib

ratio of NO₂⁺/NO⁺ from NH₄NO₃ calibrations

sLOA

sulfur-containing local organic aerosol

SO₂

sulfur dioxide

SO₄

aerosol sulfates

SOC

secondary organic carbon

SP

single particle

SusKat

sustainable atmosphere for the Kathmandu Valley

TBOA

trash burning organic aerosol

TOF-ACSM

time-of-flight aerosol chemical speciation monitor

VOC

volatile organic compound

WHO

World Health Organization

σ

standard deviation

BB

biomass burning

HR

high resolution

mAMS

mini aerosol mass spectrometer

R_measured

measured ratio of NO₂⁺/NO⁺

TB

trash burning

Data Availability Statement

Hour average time series for all species will be available at: osf.io repository linked with publication.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsearthspacechem.2c00226.

• Supporting analysis, additional measurement details with figures and tables to support the conclusions of this study (PDF)

Author Present Address

^¶ Institute for Integrated Development Studies (IIDS), Kathmandu, Nepal

Author Present Address

^◆ Laboratory for Atmospheric and Space Physics, University of Colorado at Boulder, 1234 Innovation Drive, Boulder, Colorado 80303, United States

Author Present Address

^○ Univ Paris Est Creteil and Université de Paris, CNRS, LISA, 61 Av. du Général de Gaulle, 94000 Créteil, France

Author Present Address

^∇ Aerodyne Research, Inc., 45 Manning Road, Billerica 01821, Massachusetts, United States

Author Contributions

The manuscript was written with the contributions of all authors. All authors have approved the final version of the manuscript. A.K.P., R.J.Y., E.A.S., and P.F.D. were the PIs for the project. P.F.D. and A.K.P. developed the measurement strategy. B.S.W. and M.R.G. carried out field measurements with support from K.M., P.S.P., and A.K.P. under guidance from P.F.D. M.R.I. performed formal analysis for the filters. B.S.W. performed data analysis with guidance from P.F.D. B.S.W prepared the manuscript with contributions from all co-authors.

This work was funded by NSF award numbers (AGS 1461458, AGS 1351616, AGS 1349967).

The authors declare no competing financial interest.