Seasonal variation and chemical characterization of PM2.5 in northwestern Philippines

Abstract

The seasonal and chemical characteristics of fine particulate matter (PM${}_{\mathrm{2.5}}$) were investigated in Burgos, Ilocos Norte, located at the northwestern edge of the Philippines. Each 24 h sample of fine aerosol was collected for four seasons. Fine particulate in the region shows strong seasonal variation in both concentration and composition. Highest mass concentration was seen during the boreal spring season with a mean mass concentration of 21.6 $\pm$ 6.6 $\mathrm{\mu }\mathrm{g}{\mathrm{m}}^{-\mathrm{3}}$, and lowest was in fall with a mean concentration of 8.4 $\pm$ 2.3 $\mathrm{\mu }\mathrm{g}{\mathrm{m}}^{-\mathrm{3}}$. Three-day wind back trajectory analysis of air mass reveals the influence of the northwestern Pacific monsoon regimes on PM${}_{\mathrm{2.5}}$ concentration. During southwest monsoon, sea salt was the dominant component of fine aerosols carried by moist air from the South China Sea. During northeast monsoon, on the other hand, both wind and receptor model analysis showed that higher particulate concentration was due to the long-range transport (LRT) of anthropogenic emissions from northern East Asia. Overall, sea salt and soil comprise 33 % of total PM${}_{\mathrm{2.5}}$ concentration, while local biomass burning makes up 33 %. LRT of industrial emission, solid waste burning and secondary sulfate from East Asia have a mean contribution of 34 % to the total fine particulate for the whole sampling period.

Introduction

Increasing industrial emission and open burning of biomass and solid waste have shifted much interest to the local, regional and global transport of aerosol in Asia . Aerosols are known not only for their impacts on health but also their effects on Earth's energy budget . Their influence on the climate remains as one of the main uncertainties in our understanding of the atmosphere . Rapid industrialization and urban development in the recent decades in Asia, particularly in mainland China, have led to an increase in energy consumption and consequently pollutant emissions. High emissions from the main Asian continent that are transported to other East Asian countries can contribute to elevated concentrations of ambient fine particulate elsewhere . Similarly, biomass burning from land clearing in countries such as Indonesia and Malaysia has also affected neighboring countries with reduced visibility and poor air quality . Such events can have significant social, political and economic impacts on the region.

Aside from pollutant emission, meteorology plays a significant role in transboundary pollution. Certain weather patterns create transport pathways for long-range transport (LRT) of gases and aerosols in the atmosphere. Outflow patterns of dusts and air pollutants can be induced from frontal lifting ahead of a southwestward moving cold front , or from two sequential low-pressure systems interacting with a tropical warm sector in East Asia. Also, the “warm conveyor belt” mechanism causes the seasonal uplifting and eventual transport of aerosols from East Asia to the free troposphere towards the northwest Pacific (NWP; ). In the Southeast Asian (SEA) region, biomass burning in the Maritime Continent peaks during the spring season and is modulated by multi-scale meteorological factors such as El Niño–Southern Oscillation (ENSO), inter-tropical convergence zone (ITCZ) position, Indian Ocean Dipole (IOD), Madden–Julian oscillation (MJO), monsoon winds and tropical cyclones (TCs; ), the effects of which cover large regions of SEA. In certain instances, even reaching southern China and Taiwan . The Philippines have also been identified as a source of biomass burning emissions ubiquitous in SEA. estimated an annual open field burning of 10.15 Tg of rice straw from 2002 to 2006 in the Philippines. This has led to observed elevated levels of levoglucosan and organic carbon (OC) at several sampling sites in Hong Kong during the springtime of 2004 and summer of 2006, when air parcels originating from the Pacific passed through the northern island of the Philippines . The life cycle of aerosols and their impacts on the regional climate system is the subject of several field campaigns across the region (i.e., 7-SEAS: 7-South-East Asian Studies; CAMPEX: Clouds, Aerosol, and Monsoon Processes-Philippines Experiment; YMC: Year of the Maritime Continent; BASE-ASIA: Biomass-burning Aerosols in South-East Asia: Smoke Impact Assessment; FLAMBE: Fire Locating and Monitoring of Burning Emissions).

The region's complex meteorology, warm ocean water, high sensitivity to climate change and abundant aerosol sources create a complex aerosol–cloud–climate interaction that is still not well understood . To the west of the Philippines, observed that the large-scale aerosol environment in the South China Sea (SCS) is modulated by MJO and TC activity in the NWP basin. Tropical cyclones induce significant convective activity throughout the SCS that can extend for thousands of kilometers. The associated rainfall was seen as an effective means in aerosol scavenging that leads to low aerosol concentration despite numerous sources of emission in the region. Alternatively, high aerosol concentrations were observed along western Philippines during drier periods. To the east of the Philippines lies the Pacific “warm pool”, which is among the warmest ocean areas in the world . The warm pool is also the main source of regional troposphere-to-stratospheric transported air . A study by reported the existence of a pronounced minimum in columnar ozone, as well as tropospheric column of the radical OH in the warm pool region. This will have implications on the global climate system as climate change may lead to an even warmer warm pool (Comiso et al., 2015), and at the same time likely modify the abundance of OH . These factors may contribute in prolonging the lifetime of biomass burning-induced pollutants that can increase stratospheric intrusion in the future. Moreover, monsoon wind flows that influence regional climate and weather patterns also modulate aerosol transport. Using a chemical transport model WRF-CHEM, found that the two main monsoon regimes, northeast and southwest monsoon, mostly isolate the Philippines from East Asian pollution. However, northwesterly winds that can transport pollutants from southern China can be induced by TCs during its passage to the north or northeast of the Philippines.

There is limited literature on LRT aerosol observation in the country. This is perhaps due to the geographic separation of the Philippine archipelago from the Asian continent. The lack of observations in the northern Philippine region between the East Asian subtropics and the Maritime Continent of SEA makes this location a blind spot in our knowledge of the current state of atmospheric environment. Also, satellite-based observations are hindered by persistent cloud cover over the region.

This study aims to characterize the chemical composition of PM${}_{\mathrm{2.5}}$ on the northwest region of the Philippines, identify source contributions using a receptor model and investigate existing transport pathways in the NWP region. This paper will be presented as follows: the next section will detail the characteristics of the sampling site, aerosol sampling methodology, wind back trajectory and receptor modeling. The third section will discuss the influence of the NWP/Asian monsoon on the seasonal variability of observed concentrations of fine aerosol mass and its components, as well as emission sources derived from meteorological and chemical receptor modeling. Finally, the last section will summarize the results of this study.

Methodology

Sampling site

Burgos (18.5${}^{\circ }$ N, 120.57${}^{\circ }$ E), a small town in the province of Ilocos Norte, is located in northwestern Luzon, northern Philippines as shown in Fig. . A filter-based air sampler (BGI PQ200, USA) was placed approximately 12 m above ground level atop a three-story building. The site is a rural environment surrounded by vegetation where the SCS (locally known as the West Philippine Sea) is 500 m to the west and a range of hills approximately 700 m to the east. A nearby road 100 m to the east is present, but has low daily traffic volume.

Map of East Asia, Southeast Asia and sampling site in Burgos, Philippines.

[Figure omitted. See PDF]

Burgos is classified as a Type 1 climate under the modified Coronas climate type classification where the region experiences wet season from May to September and a distinct dry season from October to April. Sampling during summer (August–September) of 2015 coincided with a monsoon break, thus all sampling days for all seasons in the present study were non-rainy days and no synoptic disturbances (i.e., TC) were present. The area is also characterized by high winds during the boreal winter and spring seasons that are mainly attributed to the cornering effect of the northeast monsoon winds to Luzon Island.

Sample collection

Daily PM${}_{\mathrm{2.5}}$ samples were collected in August to September 2015, November 2015, January to February 2016 and March 2016 to represent the boreal summer, fall, winter and spring, respectively. Two-week (14-day) sampling was conducted for each season except for the summer sampling period when the northern region of the Philippines suffered a province-wide power failure due to the effects of Typhoon Goni (locally named “Ineng”). Only 7 days of sampling was done on summer. Table 1 summarizes the sampling dates of this study. Samples were collected using a 47 mm quartz fiber filter at a flow rate of 16.7 L min${}^{-\mathrm{1}}$ from 10:00 Philippine Standard Time (PST; $+$08:00 UTC) to 10:00 PST the following day.

Chemical analysis

Prior to sampling, the quartz fiber filters are pre-heated at 900 ${}^{\circ }$C for 1.5 h to remove impurities. Each filter is then weighed before and after sampling using a microbalance (Satorius MC5). The filter is then cut into four identical parts: one for the analysis of carbonaceous components, other parts for water-soluble ionic species, for metallic elements and for the analysis of anhydrosugar.

Carbonaceous contents of PM${}_{\mathrm{2.5}}$ were measured using an elemental analyzer (Carlo Erba, model 1108). The four-part filter was divided into two, one part was heated with hot nitrogen gas (340–345 ${}^{\circ }$C) for 30 min to remove the OC fraction while the other part was analyzed without heating. The filter was then fed to the elemental analyzer to determine the amount of elemental carbon (EC) and total carbon (TC), respectively. OC concentration was calculated by getting the difference of TC and EC. Another quarter of the filter was placed in a 15 mL polyethylene (PE) bottle filled with distilled and deionized water and subjected to ultrasonic extraction for 60 min, maintained at room temperature. Ion chromatography (DIONEX DX-120) was utilized to analyze the major anions (${\mathrm{F}}^{-}$, ${\mathrm{Br}}^{-}$, ${\mathrm{Cl}}^{-}$, ${\mathrm{SO}}_{\mathrm{4}}^{\mathrm{2}-}$ and ${\mathrm{NO}}_{\mathrm{3}}^{-}$) and cations (${\mathrm{NH}}_{\mathrm{4}}^{+}$, ${\mathrm{Ca}}^{\mathrm{2}+}$, ${\mathrm{Na}}^{+}$, ${\mathrm{K}}^{+}$ and ${\mathrm{Mg}}^{\mathrm{2}+}$).

The last part of the filter was digested with a 30 mL mixed acid solution (${\mathrm{HNO}}_{\mathrm{3}}:{\mathrm{HCLO}}_{\mathrm{4}}$, $\mathrm{3}:\mathrm{7}$) at 150–200 ${}^{\circ }$C, after which the solution was diluted with 25 mL distilled and deionized water and stored in a PE bottle. Metallic elements (Al, As, Ca, Cd, Cr, Cu, Fe, K, Mg, Mn, Ni, Pb, Ti, V and Zn) were determined using an inductively coupled plasma–atomic emission spectrometer (ICP-AES, Perkin Elmer, Optima 2000DV).

Sampling dates.

Season	Sampling dates	Sampling days
Summer	27 Aug–2 Sep 2015	7
Fall	5–18 Nov 2015	14
Winter	21 Jan–3 Feb 2016	14
Spring	17–30 Mar 2016	14

Wind and receptor modeling

Analysis of wind back trajectories was done using the HYSPLIT model . Meteorological conditions were driven by output from Weather Research and Forecast (WRF) model run with 5-day spin-up time for each sampling period. WRF model with spectral nudging was used to downscale the National Centers for Environmental Prediction (NCEP) FNL (final) global reanalysis (downloaded from http://rda.ucar.edu, last access: 31 March 2017) from 1${}^{\circ }$ $\times$ 1${}^{\circ }$ horizontal spatial resolution to 15 km resolution in a two-way nested domain of 45–15 km grid resolution. Twice-daily 72 h back trajectories were then plotted and grouped into five clusters representing the general area of wind sources.

Receptor models are used to quantify the levels of air pollution, disaggregated into sources using statistical analysis of particulate matter concentrations and its chemical components. Positive matrix factorization (PMF) is a widely used receptor model by the US Environmental Protection Agency (US EPA). The US EPA PMF has been applied to identify and apportion the air pollution sources in an industrial district of the capital city of Metropolitan Manila, Philippines, in which lead (Pb) was found to make significant contributions to both coarse (PM${}_{\mathrm{2.5}-\mathrm{10}}$) and fine (PM${}_{\mathrm{2.5}}$) particulate matter fractions . PMF is utilized in the present study to identify possible emission sources of observed fine aerosols.

Wind (arrows), accumulated rainfall for each sampling period (shading, in millimeters) and wind back trajectory (red line) during (a) summer, (b) fall, (c) winter and (d) spring sampling season. The grayscale shows white for a value of 0 and black for a value of 250 in increments of 50 mm.

[Figure omitted. See PDF]

Results and discussion

Monsoon winds

The Philippines is categorized as a tropical rainforest/monsoon climate in the Köppen–Geiger climate classification. Its seasons are mainly described as wet or dry. The seasonality used in the present study mainly refers to the prevailing winds of the NWP/Asian monsoon, rather than changes in local temperature and rainfall as used in other climate classification methods. Figure 2a–d show the prevailing winds (arrow), total accumulated rainfall (shading) and twice-daily 72 h wind back trajectory (red line) in the NWP region during each of the four sampling periods. Averaged wind vectors are from the 6-hourly NCEP FNL reanalysis data and accumulated rainfall is from the TRMM 3B42A version 7 rainfall data product. Back trajectories are derived from the HYSPLIT-WRF simulations.

In the months of June–July–August (hereafter written as JJA, same for other seasons), the boreal summer season, southwest monsoon wind prevails over western Philippines as shown in Fig. a. The southwest monsoon period usually starts in the latter part of May and ends in September . The monsoon wind brings in warm moist air from the SCS, making the western coasts of the Philippines wet this season . SON (fall) in Fig. 2b is marked by the southeast propagation of the ITCZ, which results in the shifting of monsoonal winds from southwest to easterlies in September. This is followed by a shift to a northeasterly direction by the end of October or November . Figure 2c shows northeast winds prevail during DJF or boreal winter. This season is also characterized by rainfall along the eastern coastal regions of the Philippines . Lastly, in Fig. 2d, MAM (spring) marks the transition between northeast and southwest monsoon regimes. In this period, most convection stays near and south of the equator with prevailing northeast to easterly winds from the Pacific Ocean.

Daily and seasonal variation of observed PM${}_{\mathrm{2.5}}$ mass concentration.

[Figure omitted. See PDF]

The sampling periods in the present study were able to capture the climatological characteristics of each of the monsoon regimes for different boreal seasons. It is noteworthy that the monsoon transition from the northeast to southwest normally starts in the middle of March, but a late winter monsoon surge coincided with the spring sampling period. This presented a “dry” northeast monsoon that had an influence on the observed spring aerosol concentration. Moreover, the years 2015 to early 2016 were strong El Niño years; however, ENSO's influence on the NWP monsoon system will not be further discussed.

Seasonal variation of PM${}_{\mathrm{2.5}}$

The 24 h PM${}_{\mathrm{2.5}}$ mass concentration shown in Fig. 3 has strong seasonal variation. PM${}_{\mathrm{2.5}}$ mass concentration for summer, fall, winter and spring had an average value and standard deviation of 11.9 $\pm$ 5.0, 8.4 $\pm$ 2.3, 12.9 $\pm$ 4.6 and 21.6 $\pm$ 6.6 $\mathrm{\mu }\mathrm{g}{\mathrm{m}}^{-\mathrm{3}}$, respectively. The results show comparable concentrations measured from Dongsha Island in northern SCS except for heavy aerosol events previously reported in that site . However, we expect aerosol sources in northern SCS to differ from our observations based on the MODIS-derived aerosol optical thickness (AOT) analysis of . Their analysis showed northern SCS to be significantly influenced by emissions from southern and eastern China, which is not the case for northern Philippines based on our wind back trajectory simulations. Carbonaceous components EC and OC in Fig. 4 generally followed the same seasonal variation of PM${}_{\mathrm{2.5}}$. Minimum concentration was observed in fall (EC 0.40 $\pm$ 0.09; OC 0.63 $\pm$ 0.18) in $\mathrm{\mu }\mathrm{g}{\mathrm{m}}^{-\mathrm{3}}$ and maximum in spring (EC 1.03 $\pm$ 0.21; OC 1.76 $\pm$ 0.39) in $\mathrm{\mu }\mathrm{g}{\mathrm{m}}^{-\mathrm{3}}$. The annual EC and OC mean concentration and standard deviation is 0.67 $\pm$ 0.30 and 1.15 $\pm$ 0.63 $\mathrm{\mu }\mathrm{g}{\mathrm{m}}^{-\mathrm{3}}$, respectively. Measured EC likely originated from diesel buses and trucks that pass by the adjacent road. Traffic volume does not vary much near the sampling site and hence the small standard deviation observed. Overall, total carbon contribution to PM${}_{\mathrm{2.5}}$ is 13.4 % $\pm$ 3.5 %.

The mass ratio of the carbonaceous components OC $/$ EC has been previously shown to determine contribution from primary or secondary sources . EC and OC are good tracers for fossil fuel combustion and biomass burning, respectively. EC is only of primary origin while OC may be emitted directly or form from gas-to-particle conversion in the atmosphere . Accordingly, OC $/$ EC ratio is usually used in source apportionment studies of carbonaceous aerosols . Figure 5 shows the seasonal average OC $/$ EC mass ratio is 1.42, 1.74, 1.71 and 1.79 for summer, fall, winter and spring seasons, respectively. Mean OC $/$ EC ratio for all seasons are below the value of 2, which indicates that fine particulates are dominated by primary aerosol EC . On the basis of individual days, however, a third of the winter and spring data had values of OC $/$ EC greater than 2.

Daily elemental carbon (EC) and organic carbon (OC) mass concentration and OC $/$ EC ratio.

[Figure omitted. See PDF]

Southwest monsoon (summer)

Upwind regions during southwest monsoon are known to be large aerosols emitters, particularly from biomass burning. However, there is low observed aerosol concentration in this season. This is likely due to active convection around the island nations across the SCS (i.e., Borneo, Indochinese Peninsula) during the sampling period. Most air parcels indicated by the wind back trajectories originated from the marine boundary layer near these island regions. These air parcels were then transported along the western coast of northern Luzon before reaching the sampling site. The western coast of Luzon is characterized by substantial precipitation during the southwest monsoon season as a result of moist air being orographically lifted by the Cordillera mountain range in western Luzon . An average accumulated rainfall of 91.2 mm was recorded along northwestern Luzon coast throughout the 7-day summer sampling period. These factors would have resulted in the suppression of biomass burning in the SEA region and scavenging of particulates along the path of the transported air parcels. In addition, the WRF simulation used to drive the HYSPLIT model showed a strong diurnal cycle of land–sea breeze along the western Luzon coasts. Nighttime land breeze carries polluted air from central and southwestern Luzon northwestward to the SCS through Lingayen Gulf, after which daytime sea breeze pushes back these polluted air masses inland along the northwest Luzon region. A similar land–sea breeze influence on the back-and-forth spatial distribution of pollutants was observed by along the coastal region of southern Taiwan.

Scatter plot of OC and EC for summer, fall, winter and spring sampling period (in $\mathrm{\mu }\mathrm{g}{\mathrm{m}}^{-\mathrm{3}}$). Bold dashed line indicates OC $/$ EC ratio of 2.

[Figure omitted. See PDF]

Southwest to northeast monsoon transition (fall)

November (fall) sampling showed the lowest mass concentrations among all seasons. During the fall monsoonal transition regime, easterlies bring in air masses from the northwest Pacific Ocean (shown in Fig. 2b), where no known large emission sources are present. Contribution from the eastern region of northern Philippines appears to be minimal as the northwest and northeast Philippines are separated by the northern hills of the Cordillera range. Also, northeast Philippines is mainly composed of agricultural land with only little to moderate urban activities.

Cluster analysis of 72 h HYSPLIT-WRF wind back trajectories arriving every 12 h for each sampling day for all seasons. Color represents each of the five clusters generated by the analysis.

[Figure omitted. See PDF]

Northeast monsoon (winter and spring)

Highest mass concentration is seen during springtime, followed by the winter observation. Strong northeasterly wind affected both sampling periods. Wind back trajectories of both seasons in Fig. 2c and d show air parcels come from northern East Asia. However, better outflow patterns of pollutants from northern Asia during springtime may have contributed to higher observed mass concentrations in March. In addition, heavier precipitation from the Meiyu/Baiu front located along the East Asian subtropics during winter, as seen in Fig. 2c, likely reduced the transported aerosols by wet scavenging before reaching the Philippines.

Wind trajectory cluster analysis

Figure 6 shows all the wind back trajectories grouped into five clusters. Of all simulated wind back trajectories, 17 % (pink) occurred during southwest monsoon showed that air parcels originated from the SCS and moved along the western coast of northern Luzon. In fall, 27 % (blue) and 14 % (red) originated from the near and far east Pacific, respectively. Total of 41 % of pristine air parcels originated from the Pacific waters. This period is when the lowest PM${}_{\mathrm{2.5}}$ mass concentration including most of its components was observed. During northeast monsoon, 30 % comes from northern East Asia and another 13 % from East China sea. All five back trajectory clusters were simulated from ground-to-ground transport. The chemical characteristics and possible aerosol sources will be discussed in the following sections.

Seasonal mean mass concentration and standard deviation of (a) PM${}_{\mathrm{2.5}}$ and its ionic components, (b) ${\mathrm{NO}}_{\mathrm{3}}^{-}$, (c) ${\mathrm{SO}}_{\mathrm{4}}^{\mathrm{2}-}$, (d) ${\mathrm{NH}}_{\mathrm{4}}^{+}$, (e) ${\mathrm{Na}}^{+}$, (f) ${\mathrm{Cl}}^{-}$, (g) ${\mathrm{K}}^{+}$, (h) ${\mathrm{Ca}}^{\mathrm{2}+}$ and (i) ${\mathrm{Mg}}^{\mathrm{2}+}$ (unit in $\mathrm{\mu }\mathrm{g}{\mathrm{m}}^{-\mathrm{3}}$).

[Figure omitted. See PDF]

Ionic and metallic components

Figure 7 shows the seasonal mean and standard deviation of PM${}_{\mathrm{2.5}}$ and some water-soluble ionic components. It is apparent that ${\mathrm{NO}}_{\mathrm{3}}^{-}$, ${\mathrm{SO}}_{\mathrm{4}}^{\mathrm{2}-}$ and ${\mathrm{NH}}_{\mathrm{4}}^{+}$ shown in Fig. 7b, c and d, respectively, also follow the seasonal variation of PM${}_{\mathrm{2.5}}$ mass concentration. Minimum concentration was observed in fall and maximum during spring sampling. These three components are associated with secondary inorganic aerosol and make up on average 69 % $\pm$ 4 % of the total water-soluble ions. Among all ionic species, ${\mathrm{SO}}_{\mathrm{4}}^{\mathrm{2}-}$ has the highest contribution at 2.43 $\pm$ 0.37 $\mathrm{\mu }\mathrm{g}{\mathrm{m}}^{-\mathrm{3}}$, followed by ${\mathrm{NO}}_{\mathrm{3}}^{-}$ at 0.98 $\pm$ 0.25 $\mathrm{\mu }\mathrm{g}{\mathrm{m}}^{-\mathrm{3}}$ and ${\mathrm{NH}}_{\mathrm{4}}^{+}$ at 0.71 $\pm$ 0.14 $\mathrm{\mu }\mathrm{g}{\mathrm{m}}^{-\mathrm{3}}$. Seasonality of ${\mathrm{Na}}^{+}$ with mean concentration of 0.37 $\pm$ 0.06 $\mathrm{\mu }\mathrm{g}{\mathrm{m}}^{-\mathrm{3}}$ and ${\mathrm{Ca}}^{\mathrm{2}+}$ (0.34 $\pm$ 0.06 $\mathrm{\mu }\mathrm{g}{\mathrm{m}}^{-\mathrm{3}}$) also shows the same seasonal variability. However, this seasonality is not apparent for the ions ${\mathrm{Cl}}^{-}$, ${\mathrm{K}}^{+}$ and ${\mathrm{Mg}}^{\mathrm{2}+}$.

Summary of mass concentration of PM${}_{\mathrm{2.5}}$ and its components (unit is $\mathrm{\mu }\mathrm{g}{\mathrm{m}}^{-\mathrm{3}}$).

	Summer	Fall	Winter	Spring
PM${}_{\mathrm{2.5}}$	11.9 $\pm$ 4.6	8.4 $\pm$ 2.2	14.2 $\pm$ 7.3	21.6 $\pm$ 6.6
EC	0.50 $\pm$ 0.09	0.40 $\pm$ 0.09	0.67 $\pm$ 0.19	1.03 $\pm$ 0.20
OC	0.93 $\pm$ 0.61	0.63 $\pm$ 0.17	1.18 $\pm$ 0.62	1.76 $\pm$ 0.37
${\mathrm{NO}}_{\mathrm{3}}^{-}$	0.71 $\pm$ 0.21	0.60 $\pm$ 0.16	0.80 $\pm$ 0.27	1.81 $\pm$ 1.29
${\mathrm{SO}}_{\mathrm{4}}^{\mathrm{2}-}$	1.69 $\pm$ 0.43	1.50 $\pm$ 0.42	2.48 $\pm$ 0.94	4.06 $\pm$ 1.67
${\mathrm{NH}}_{\mathrm{4}}^{+}$	0.53 $\pm$ 0.15	0.32 $\pm$ 0.09	0.63 $\pm$ 0.21	1.35 $\pm$ 0.72
${\mathrm{Na}}^{+}$	0.37 $\pm$ 0.11	0.31 $\pm$ 0.08	0.37 $\pm$ 0.15	0.44 $\pm$ 0.18
${\mathrm{Cl}}^{-}$	0.69 $\pm$ 0.17	0.52 $\pm$ 0.17	0.57 $\pm$ 0.23	0.60 $\pm$ 0.24
${\mathrm{K}}^{+}$	0.07 $\pm$ 0.03	0.08 $\pm$ 0.02	0.24 $\pm$ 0.07	0.53 $\pm$ 0.38
${\mathrm{Ca}}^{\mathrm{2}+}$	0.25 $\pm$ 0.05	0.21 $\pm$ 0.05	0.25 $\pm$ 0.10	0.63 $\pm$ 0.26
${\mathrm{Mg}}^{\mathrm{2}+}$	0.20 $\pm$ 0.04	0.14 $\pm$ 0.03	0.10 $\pm$ 0.02	0.29 $\pm$ 0.07
Al	0.05 $\pm$ 0.00	0.05 $\pm$ 0.01	0.07 $\pm$ 0.02	0.12 $\pm$ 0.04
Fe	0.12 $\pm$ 0.00	0.10 $\pm$ 0.02	0.15 $\pm$ 0.02	0.17 $\pm$ 0.02
Cd	0.01 $\pm$ 0.00	0.03 $\pm$ 0.00	0.04 $\pm$ 0.01	0.13 $\pm$ 0.04
Cr	0.02 $\pm$ 0.00	0.01 $\pm$ 0.00	0.05 $\pm$ 0.01	0.10 $\pm$ 0.01
Ni	0.00 $\pm$ 0.00	0.01 $\pm$ 0.00	0.03 $\pm$ 0.01	0.09 $\pm$ 0.02
Pb	0.06 $\pm$ 0.01	0.06 $\pm$ 0.01	0.09 $\pm$ 0.02	0.11 $\pm$ 0.01
Levoglucosan*	0.63 $\pm$ 0.86	1.4 $\pm$ 0.56	6.0 $\pm$ 2.4	19.1 $\pm$ 2.8

${}^{*}$ unit in ng m${}^{-\mathrm{3}}$.

Seasonal mean mass concentration and standard deviation of metallic components (a) Al, (b) Cd, (c) Cr, (d) Fe, (e) Ni and (f) Pb (unit in $\mathrm{\mu }\mathrm{g}{\mathrm{m}}^{-\mathrm{3}}$).

[Figure omitted. See PDF]

In Fig. 7f, summer chlorine (${\mathrm{Cl}}^{-}$) sampling shows highest concentration of 0.69 $\pm$ 0.17 $\mathrm{\mu }\mathrm{g}{\mathrm{m}}^{-\mathrm{3}}$ and the rest of the seasons with nearly constant concentration of 0.52 $\pm$ 0.09 $\mathrm{\mu }\mathrm{g}{\mathrm{m}}^{-\mathrm{3}}$ for fall, 0.57 $\pm$ 0.07 $\mathrm{\mu }\mathrm{g}{\mathrm{m}}^{-\mathrm{3}}$ for winter and 0.60 $\pm$ 0.12 $\mathrm{\mu }\mathrm{g}{\mathrm{m}}^{-\mathrm{3}}$ for spring. We attribute the high ${\mathrm{Cl}}^{-}$ content in summer to sea salt carried by the southwest monsoon wind. Potassium (${\mathrm{K}}^{+}$) has mean concentration of 0.23 $\pm$ 0.06 $\mathrm{\mu }\mathrm{g}{\mathrm{m}}^{-\mathrm{3}}$. ${\mathrm{K}}^{+}$ is a relatively abundant element in crustal rocks and is also used as a tracer for wood burning due to its significant amount in wood biomass . Compared with levoglucosan measurements shown in Table 2, ${\mathrm{K}}^{+}$ is highly correlated with levoglucosan with correlation coefficient $r=\mathrm{0.99}$ at 95 % confidence interval ($p<\mathrm{0.05}$). This indicates that measured ${\mathrm{K}}^{+}$ is mainly from open burning of biomass, which is more widespread during the dry season of winter and spring. Magnesium (${\mathrm{Mg}}^{\mathrm{2}+}$) has maximum concentration in spring and lowest in winter. For summer and fall, ${\mathrm{Mg}}^{\mathrm{2}+}$ is highly correlated with ${\mathrm{Ca}}^{\mathrm{2}+}$ (summer $r=\mathrm{0.94}$; fall $r=\mathrm{0.97}$, both at $p<\mathrm{0.05}$), on the other hand, no significant correlations were found in other seasons. This suggests that the source of ${\mathrm{Mg}}^{\mathrm{2}+}$ is mostly from mineral dust (carbonate mineral) when highly correlated with ${\mathrm{Ca}}^{\mathrm{2}+}$ . The high concentration of ${\mathrm{Mg}}^{\mathrm{2}+}$ in spring is therefore attributed to non-local sources. Source attribution for these ions is further discussed in the results of receptor modeling in the next section.

Figure 8 shows some metallic components of measured fine particulates. Metallic components Al and Fe in Fig. 8a and d, respectively, are associated with crustal origins . Seasonal variations of heavy metals Cd, Cr, Ni and Pb are shown in Fig. 8b, c, e and f, respectively. Heavy metal components of fine particulates pose a health risk . Particularly, Ni, Cd and Cr are identified as human carcinogens while Pb is toxic and exposure can lead to permanent adverse health effects in humans . Dispersion of metals embedded in particulates also determines the rate at which metals deposit on Earth's surface . All heavy metal components were evidently high in spring followed by the winter sampling period. Ambient concentrations of anthropogenic components depend on distance from source location and transport process . Since no large industries or power plants are present within 250 km of the sampling site, these toxic components likely originated from upwind regions during northeast monsoon. No significant correlations were found between these metallic components with ionic components associated with secondary inorganic aerosols. This suggests that these heavy metal components come from several different sources. Table 2 is the summary of the seasonal mean mass concentration and their corresponding standard deviation of PM${}_{\mathrm{2.5}}$ and its components, including the anhydrosugar levoglucosan.

Figure 9a shows the ratio of cation and anion close to unity for all seasons (Fig. 9 is in units of equivalent concentrations). This indicates good charge balance of atmospheric aerosols and high data quality used in the current study. Figure 9b shows the scatter plot of ${\mathrm{Cl}}^{-}$ vs. ${\mathrm{Na}}^{+}$. The ratio ${\mathrm{Cl}}^{-}$ $/$ ${\mathrm{Na}}^{+}$ shows highest value in summer with 1.19, 1.08 for fall, 0.99 for winter and 0.89 for spring. The ratio indicates that summer ${\mathrm{Cl}}^{-}$ is mainly from sea salt, where mean ${\mathrm{Cl}}^{-}$ $/$ ${\mathrm{Na}}^{+}$ ratio of sea salt is equivalent to 1.17 . This result supports our initial hypothesis that the high summer ${\mathrm{Cl}}^{-}$ concentration mainly comes from sea salt. In addition, there may be ${\mathrm{Cl}}^{-}$ depletion in the rest of the seasons due to the following factors: (1) farther distance from upwind coast , (2) high sulfate and nitrate concentration during northeast monsoon may have reacted with ${\mathrm{Cl}}^{-}$ in sea-salt-forming gas-phase HCl in the process and (3) excess ${\mathrm{Na}}^{+}$ may have come from resuspended soil due to stronger wind in non-summer seasons. This is further supported by the high correlation values found between Fe and Al in winter ($r=\mathrm{0.71}$ at $p<\mathrm{0.05}$) and spring ($r=\mathrm{0.88}$ at $p<\mathrm{0.05}$ ) measurements, suggesting higher loading of uplifted dust blown by strong winds during those seasons.

Also mentioned in the previous section, ${\mathrm{Mg}}^{\mathrm{2}+}$ is highly correlated with ${\mathrm{Ca}}^{\mathrm{2}+}$ for summer and fall. This indicates mineral dust as main source of ${\mathrm{Mg}}^{\mathrm{2}+}$ in these seasons. In terms of the ratio of ${\mathrm{Mg}}^{\mathrm{2}+}$ $/$ ${\mathrm{Na}}^{+}$, among all seasons, winter shows closest to the mean sea salt ratio of 0.23 (Chester, 1990), indicating mostly non-sea-salt source for Mg except winter. Furthermore, ratio of both ${\mathrm{Mg}}^{\mathrm{2}+}$ $/$ ${\mathrm{Ca}}^{\mathrm{2}+}$ and ${\mathrm{Mg}}^{\mathrm{2}+}$ $/$ ${\mathrm{Na}}^{+}$ tends to vary more and spread out during spring season sampling as seen in graphs of Fig. 9c and d. For the components associated with secondary inorganic aerosols, Fig. 9e shows the ratio of ${\mathrm{NH}}_{\mathrm{4}}^{+}$ $/$ ${\mathrm{SO}}_{\mathrm{4}}^{\mathrm{2}-}$ all below unity (bold dashed line). The ratio points to ${\mathrm{NH}}_{\mathrm{4}}^{+}$ not fully neutralizing ${\mathrm{SO}}_{\mathrm{4}}^{\mathrm{2}-}$ all throughout the sampling periods. This is possibly due to nearby sources of sea salt sulfate (SS ${\mathrm{SO}}_{\mathrm{4}}^{\mathrm{2}-}$) in the region. Similarly, the ratio ${\mathrm{NH}}_{\mathrm{4}}^{+}$ $/$ [${\mathrm{NO}}_{\mathrm{3}}^{-}+{\mathrm{SO}}_{\mathrm{4}}^{\mathrm{2}-}$] is also found to be below unity for all seasons. This suggests the presence of ${\mathrm{NH}}_{\mathrm{4}}^{+}$${\mathrm{NO}}_{\mathrm{3}}^{-}$ as well as other forms of ${\mathrm{NO}}_{\mathrm{3}}^{-}$ in the region.

Source contribution

The US EPA PMF 5.0 was used to resolve the contribution of the identified factor sources to the PM${}_{\mathrm{2.5}}$ concentration on each sampling day. The US EPA 5.0 uses a weighted least squares model, weighted based on known uncertainty or error of the elements of the data matrix . The goal is to obtain the minimum $Q$ value after several iterations, keeping the residuals at the most reasonable levels and having a sensible and rational factor profile. Details of the US EPA 5.0 are described elsewhere . Here, all 49 sampling data sets were used to resolve the factor and contribution profiles of PM${}_{\mathrm{2.5}}$ in northwestern Philippines. An extra 10 % modeling uncertainty was added to the data to obtain the optimum convergence of the $Q$ value and acceptable scaled residuals in the runs.

Scatter plot of (a) total anion and cation, (b) ${\mathrm{Na}}^{+}$ and ${\mathrm{Cl}}^{-}$, (c) ${\mathrm{Mg}}^{\mathrm{2}+}$ and ${\mathrm{Na}}^{+}$, (d) ${\mathrm{Mg}}^{\mathrm{2}+}$ and ${\mathrm{Ca}}^{\mathrm{2}+}$, (e) ${\mathrm{NH}}_{\mathrm{4}}^{+}$ and ${\mathrm{SO}}_{\mathrm{4}}^{\mathrm{2}-}$ and (f) ${\mathrm{NH}}_{\mathrm{4}}^{+}$ and ${\mathrm{NO}}_{\mathrm{3}}^{-}$ $+$ ${\mathrm{SO}}_{\mathrm{4}}^{\mathrm{2}-}$ (unit in equivalent concentration).

[Figure omitted. See PDF]

Source factor profiles from PMF analysis: (a) sea salt, (b) resuspended fine dust, (c) local, solid waste burning, (d) LRT, industrial emissions, (e) LRT, secondary sulfate and (f) LRT, solid waste burning.

[Figure omitted. See PDF]

Summary of seasonal and annual source profile contribution (unit is $\mathrm{\mu }\mathrm{g}{\mathrm{m}}^{-\mathrm{3}}$).

Source	Summer	Fall	Winter	Spring	Annual
Sea salt	5.7 $\pm$ 1.5 (49 %)	3.4 $\pm$ 1.4 (40 %)	1.9 $\pm$ 1.1 (14 %)	3.3 $\pm$ 2.6 (15 %)	3.3 $\pm$ 2.1 (23 %)
Resuspended fine dust	0.4 $\pm$ 0.4 (3 %)	1.2 $\pm$ 0.4 (11 %)	2.1 $\pm$ 0.7 (15 %)	1.5 $\pm$ 1.2 (7 %)	1.4 $\pm$ 1.0 (10 %)
Local solid waste burning	5.0 $\pm$ 1.3 (43 %)	2.8 $\pm$ 1.1 (33 %)	4.8 $\pm$ 3.9 (35 %)	6.2 $\pm$ 5.1 (29 %)	4.7 $\pm$ 3.7 (33 %)
LRT, industrial emission	0.02 $\pm$ 0.03 (0.1 %)	0.0 $\pm$ 0.0 (0 %)	3.1 $\pm$ 2.7 (22 %)	4.0 $\pm$ 3.8 (18 %)	2.0 $\pm$ 3.0 (14 %)
LRT, secondary sulfate	0.1 $\pm$ 0.1 (1 %)	0.1 $\pm$ 0.1 (1 %)	1.2 $\pm$ 0.4 (8 %)	3.3 $\pm$ 2.0 (15 %)	1.3 $\pm$ 1.7 (9 %)
LRT, solid waste burning	0.4 $\pm$ 0.3 (3 %)	1.0 $\pm$ 0.3 (11 %)	0.7 $\pm$ 0.5 (5 %)	3.5 $\pm$ 2.8 (16 %)	1.5 $\pm$ 2.0 (11 %)

Enrichment factors.

	Solid waste burning
	local	LRT
${\mathrm{K}}^{+}$ $/$ OC	0.4	0.2
${\mathrm{K}}^{+}$ $/$ EC	0.9	0.4
${\mathrm{K}}^{+}$ $/$ Zn	3.7	2.7
${\mathrm{K}}^{+}$ $/$ ${\mathrm{NO}}_{\mathrm{3}}^{-}$	3.6	0.1
	LRT episodes
	secondary sulfate	industrial	resuspended fine dust
(${\mathrm{Ca}}^{\mathrm{2}+}$ $+$ ${\mathrm{Mg}}^{\mathrm{2}+}$) $/$ ${\mathrm{Na}}^{+}$	5.7	1.6	null
${\mathrm{NO}}_{\mathrm{3}}^{-}$ $/$ ${\mathrm{SO}}_{\mathrm{4}}^{\mathrm{2}-}$	null	0.9	0.1

Daily contribution for each source profile: (a) sea salt; (b) resuspended fine dust; (c) local, solid waste burning; (d) LRT, industrial emissions; (e) LRT, secondary sulfate; and (f) LRT, solid waste burning for all sampling periods (unit is $\mathrm{\mu }\mathrm{g}{\mathrm{m}}^{-\mathrm{3}}$).

[Figure omitted. See PDF]

Here, we obtained six source factors, namely (1) sea salt, (2) resuspended fine dust, (3) local solid waste burning, (4) LRT of industrial emissions, (5) LRT solid waste burning and (6) LRT secondary sulfate. Figure 10 shows the profiles of the factors (sources) identified. Figure 11 shows the daily contribution per season for each of the source profiles. Using the source contributions, we were able to resolve the seasonal concentration of the sources, consistent with the factor profiles and fingerprints. For instance, elevated levels of sea salt contributed mainly during the summer season (5.7 $\pm$ 1.5 $\mathrm{\mu }\mathrm{g}{\mathrm{m}}^{-\mathrm{3}}$), consistent with our chemical analysis. On the other hand, LRT industrial emissions are observed at elevated levels during spring (4.0 $\pm$ 3.8 $\mathrm{\mu }\mathrm{g}{\mathrm{m}}^{-\mathrm{3}}$) and winter seasons (3.1 $\pm$ 2.7 $\mathrm{\mu }\mathrm{g}{\mathrm{m}}^{-\mathrm{3}}$), consistent with the chemical and wind back trajectory analysis discussed above. Table 3 summarizes the seasonal and total contribution of source factors to fine particulate matter of the region. Overall, natural primary sources sea salt and resuspended fine dust constitute 33 % of atmospheric aerosols. Another 33 % or one-third is due to local solid waste burning. This includes open burning of biomass in the dry season for the purpose of land clearing ubiquitous in SEA. Lastly, 34 % is due to LRT sources from industrial emission, solid waste burning and secondary sulfate.

Enrichment factor

Analysis of the enrichment factor is done to further characterize the composition and associations of the chemical components of PM${}_{\mathrm{2.5}}$. The analysis relates the concentration of PM${}_{\mathrm{2.5}}$ components that are known to be anthropogenic to those that are found stable in the crust, or those that are naturally found in the local atmosphere .

The enrichment factors of the chemical markers of identified sources are tabulated in Table 4. Factors associated with solid waste burning are divided into local and LRT burning factors. Both have high associations with ${\mathrm{K}}^{+}$, Zn and OC. The LRT solid waste burning factor exhibits strong association with ${\mathrm{NO}}_{\mathrm{3}}^{-}$, ${\mathrm{NH}}_{\mathrm{4}}^{+}$, ${\mathrm{Mg}}^{\mathrm{2}+}$, ${\mathrm{Ca}}^{\mathrm{2}+}$, ${\mathrm{K}}^{+}$, Zn, OC and EC. The enrichment factors of OC, EC, Zn and ${\mathrm{NO}}_{\mathrm{3}}^{-}$ with respect to ${\mathrm{K}}^{+}$ for local burning decreased to 50 % when compared to the LRT counterpart, indicating the decrease in ageing of the PM${}_{\mathrm{2.5}}$ components as particles are transported over a long distance. The two other LRT factors identified, secondary sulfate and industrial factor source, showed strong associations with the heavy metals Cr, Ni, Cu, Cd and Pb. These chemical markers are reported in petroleum, chemical and manufacturing industries that are not locally present. The secondary sulfate source marked an enrichment factor for (${\mathrm{Ca}}^{\mathrm{2}+}$ $+$ ${\mathrm{Mg}}^{\mathrm{2}+}$) $/$ ${\mathrm{Na}}^{+}$ of 5.7, which is about the same value as the enrichment factor of a certified reference material of China loess soil , while that of the industrial emission factor (1.6) corresponds to the enrichment factor of sea salt ageing on processed dust particles from a marine background site in Korea .

It is noteworthy that significant contribution from long distance sources are observed during the northeast monsoon seasons of winter and spring. Analysis of wind back trajectory, PMF model and chemical components all demonstrate the existence of transboundary aerosols by way of the northeast monsoon wind. Furthermore, relatively lower concentration of components linked to LRT found in winter is likely modulated by rainfall associated with the Meiyu/Baiu front (shown in Fig. 2c). More (less) frontal rain in winter (spring) resulted in increased (decreased) aerosol scavenging, which affected the overall transport flow of LRT fine aerosols. Figure 12 shows a high correlation ($r=\mathrm{0.87}$) between the observed and reconstructed PMF-modeled PM${}_{\mathrm{2.5}}$ mass concentration, providing high confidence on the PMF analysis.

Conclusions

This study has described the seasonal characteristics of fine particulates (PM${}_{\mathrm{2.5}}$) in Burgos, Ilocos Norte, located in the northwestern edge of the Philippines. This region is located between the East Asian subtropics and the Maritime Continent. Both regions are known emitters of large quantities of anthropogenic aerosols. Observed fine particulates are mainly modulated by the NWP monsoon winds. PM${}_{\mathrm{2.5}}$ shows strong seasonality where the lowest mean concentration is found during fall season when easterly winds prevail. High concentrations were found in winter and springtime during the northeast monsoon season. PM${}_{\mathrm{2.5}}$ mass concentration for summer, fall, winter and spring had an average value and standard deviation of 11.9 $\pm$ 5.0, 8.4 $\pm$ 2.3, 12.9 $\pm$ 4.6 and 21.6 $\pm$ 6.6 $\mathrm{\mu }\mathrm{g}{\mathrm{m}}^{-\mathrm{3}}$, respectively.

Measured vs. PMF model reconstruction of PM${}_{\mathrm{2.5}}$ mass concentration (unit is $\mathrm{\mu }\mathrm{g}{\mathrm{m}}^{-\mathrm{3}}$).

[Figure omitted. See PDF]

Components of fine particulates also showed distinct seasonality. Carbonaceous aerosol components EC and OC have an annual mean value of 0.67 $\pm$ 0.30 and 1.15 $\pm$ 0.63 $\mathrm{\mu }\mathrm{g}{\mathrm{m}}^{-\mathrm{3}}$, respectively, both lowest in fall and highest in spring. EC and OC collectively make up 13.4 % $\pm$ 3.5 % of observed fine particulates. The ionic and metallic components of fine aerosols also varied by sampling period which generally followed the seasonal variation of PM${}_{\mathrm{2.5}}$, and make up 44.4 % $\pm$ 10.1 % and 11.7 % $\pm$ 3.8 % of PM${}_{\mathrm{2.5}}$ mass, respectively. Analysis of the chemical components reveals high sea salt content in summer when southwest monsoon winds prevail, and high concentration of components associated with secondary inorganic aerosols (i.e., ${\mathrm{NO}}_{\mathrm{3}}^{-}$, ${\mathrm{SO}}_{\mathrm{4}}^{\mathrm{2}-}$ and ${\mathrm{NH}}_{\mathrm{4}}^{+}$) as well as anthropogenic pollutants (i.e., heavy/toxic metals) during northeast monsoon. HYSPLIT-WRF wind back trajectory results show air masses originating from East Asia move along the northeasterly wind in winter and spring seasons. Winter sampling showed comparatively lower concentrations of PM${}_{\mathrm{2.5}}$ than spring. We attribute this to the scavenging of transported aerosols by the Meiyu/Baiu front, which had higher precipitation during the winter sampling period.

Positive matrix factorization (PMF) of the US EPA was used to determine the source contributors of fine particulate in the region. The results of the PMF receptor model and wind analysis were consistent and complementary. Here, six source profiles were obtained using the receptor model, namely (1) sea salt, (2) resuspended fine dust, (3) local solid waste burning, (4) LRT of industrial emissions, (5) LRT solid waste burning and (6) LRT secondary sulfate. Consistent with the chemical analysis, high sea salt in summer contributes to almost half of aerosol content for that season. Resuspended fine dust is seen to increase in the spring and winter season when strong winds prevail over the sampling site. Open burning of biomass and solid waste is widespread in the dry seasons of winter and spring. This is seen in the seasonality of ${\mathrm{K}}^{+}$ and the anhydrosugar levoglucosan, which were found to be highly correlated with one another. LRT of anthropogenic fine particulates were observed during winter and springtime when the northeast monsoon serves as transport pathway for East Asian aerosols to reach the northern part of the Philippines. The annual mean source contribution of transboundary industrial emission, secondary sulfate and solid waste burning was 14, 9 and 11 %, respectively. In total, LRT contributes to one-third of aerosol content in the region.

To our knowledge, this is the first comprehensive analysis of aerosol characteristics in this region of the Philippines. Also, this is the first study to confirm long-range transport of East Asian aerosols to the country. It would be interesting to see its implications on the region's radiative forcing, aerosol–cloud–climate interaction and stratospheric intrusion, if there are any. These are questions essential to better understanding the region's atmosphere.

The measurement and simulation data presented in the paper are available from the authors upon request ([email protected]).

The authors declare that they have no conflict of interest.

Acknowledgements

The authors would like to acknowledge the Department of Science and Technology (Philippines) and the Ministry of Science and Technology (Taiwan ROC) for funding the project entitled “Tempospatial Distribution and Transboundary Transport of Atmospheric Fine Particles Across Bashi Channel, Taiwan Strait and South China Sea”. Edited by: Leiming Zhang Reviewed by: two anonymous referees