1. Introduction
Antarctica, known as the white continent, is a remote region interconnected with the rest of the world by oceanic and atmospheric couplings, which plays a dominant role in the climatic system. The thermohaline circulation, responsible for the large-scale heat transport in the Ocean, is deeply modulated in the Southern Hemisphere by the seasonal sea ice sheet formations in the Antarctic. The prevailing strong westerly winds surrounding the Antarctic near surface and the stratospheric polar vortex formed in spring are atmospheric patterns responsible for the oceanic Antarctic circumpolar current. These two examples alone underline the importance of the Antarctic for global climate. For these reasons, among others, there is an increasing historical interest in scientific research on the Antarctic region in addition to international geopolitical agreements. The fate of its ice sheets plays a dominant role in climate regulation (Luis, 2013; Kennicutt et al., 2019).
Antarctica is one of the most pristine and unexplored regions of the world. However, there is evidence that it is a sink for atmospheric aerosols transported from remote locations (Bargagli, 2016; Anzano et al., 2022), negatively affecting the ecosystems and environmental conditions of this continent (Marina-Montes et al., 2020). Depending on the wind and meteorological conditions, atmospheric aerosols can be transported thousands of kilometers in days (Zhong et al., 2019; Jia and Zhang, 2020; Xiong et al., 2020). This transport undergoes microphysical and chemical transformations, making it spatially and temporally non-uniform (Bellouin et al., 2020). Aerosol concentration, size distribution, chemical composition, and dynamic behavior in the atmosphere play a crucial role in governing radiative transfer and, thus, the climate system (Bellouin et al., 2020; Marina-Montes et al., 2020).
King George Island (KGI) is the largest of the South Shetland Islands. It is located on the Antarctic Peninsula and contains many research stations (Polyakov et al., 2020). There are several reports about atmospheric aerosols conducted in the Antarctic Peninsula (Artaxo et al., 1992; Ángeles et al., 2020; Hong et al., 2020; Polyakov et al., 2020; Fan et al., 2021; Szumińska et al., 2021; Gonçalves et al., 2022). However, despite progress in characterizing and determining the origin of atmospheric aerosols (Marina-Montes et al., 2020), there is still limited information on the source areas of air masses that move particles toward the white continent.
The Peruvian Antarctic Scientific Station Machu Picchu (Estación Científica Antártica Machu Picchu, ECAMP) is located in the KGI, at the northern end of the Antarctic Peninsula, very close to the open ocean with maritime climatic conditions. ECAMP only operates during the Austral summer, covering periods from December to March at most (Ángeles et al., 2020), so studies on particles and their source areas at the site are still insufficient.
Therefore, this research aims to characterize the chemical composition of aerosol monitors and atmospheric particles at ECAMP, analyze the contribution of their emission sources, and investigate the origin of the air masses that contribute to their transport to the site during the austral summer of 2020. In this regard, this study provides the first valuable data set on trace elements and atmospheric particles in the understudied area of ECAMP. This dataset and reported results will contribute to filling a gap in the knowledge about aerosol behavior in ECAMP.
2. Materials and methods
2.1 Study site
The study site where the measurement and collection equipment were installed was the ECAMP (62.09º S, 58.47º W) in King George Island, north of the Antarctic Peninsula (Fig. 1). The sampling site was about 1 km east of the ECAMP’s main building and about 20 m from the sea coast. The mean annual temperature is -2.8 ºC with an annual precipitation of 729 mm. There are about 100 days of sunshine (Polyakov et al., 2020). The ECAMP is located south of Admiralty Bay (Mackellar Inlet) below the Machu Pichu Mountain, about 600 m from the Znosko Glacier. It has gravelly and sandy soils, with rocks and flood zones resulting from snow thawing.
2.2 Measurements
The sampling periods were between February 7 and 23, 2020, during the Peruvian Antarctic Campaign ANTAR XXVII organized by the Ministry of Foreign Affairs of Peru. PM2.5 and PM10 were collected simultaneously on a Teflon filter (diameter of 47 mm, pore size of 2.0 µm, Whatman, USA) using two Partisol samplers (model 2000-FRM, Thermo Scientific, USA). The first sampler was installed with a stainless steel 2.5 μm cut point cyclone, and the other with an inlet-type sampling head to cut at 10 µm (US-EPA). The flow rate was 16.7 L/min. A total of five samples were collected (February 7-9, 10-13, 14-17, 17-20, and 20-23). The collection period was set at 72 h following the recommendation of Artaxo et al (1992), to ensure sufficient material was available for detecting trace elements in the filters. The filters with samples were stored and transported to the laboratory in Peru in labeled containers at 4 ºC.
The sampling of atmospheric aerosols was complemented with aerosol size distribution measurements and black carbon (BC) concentration from February 4 to 23, 2020. The counting data were collected by a laser diode-based optical particle counter with eight channels (model 212, Met One Instruments, USA). The size channel ranges were 0.3; >0.3-0.5; >0.5-0.7; >0.7-1.0; >1.0-2.0; >2.0-3.0; >3.0-5.0, and >5.0-10.0 μm. The channels will be named from now on the text as PM03, PM0.3-0.5, PM0.5-0.7, PM0.7-1, PM1-2, PM2-3, PM3-5, and PM5-10, respectively. The instruments continuously sampled air at 1 L/min and provided data count per minute for each channel.
An Aethalometer (model AE33, Magee Scientific, USA) was used to measure the BC concentration. It is based on the change of absorbance of light through the pass of a filter continuously loaded with aerosols. The passing light is analyzed at seven wavelengths: 370, 470, 520, 590, 660, 880, and 950 nm. A 2.5 μm cut point cyclone was placed at the inlet to avoid coarse particles. Particles in the incoming airflow are deposited on the polytetrafluoroethylene (PTFE) filter. The instrument operates continuously at a flow rate of 5 L/min (Drinovec et al., 2015; Suárez et al., 2017).
The suction tubes of the installed instruments and Partisol samplers were mounted at 1 m from the ceiling level of the support platform and approximately 3 m above the ground level to avoid surface effects and soil contamination.
Meteorological data was obtained from the automatic meteorological station (Model Vaisala MAWS-301) installed at the Peruvian station by the Directorate of Aeronautical Meteorology of Peru (Dirección de Meteorología Aeronáutica, DIRMA) during the study. This equipment was installed 4 m north of the sampling site.
2.3 Gravimetric analysis
PM10 and PM2.5 mass concentrations were determined by the gravimetric method using Teflon filters in the Partisol samplers. The filters were weighed before and after sampling using an electronic microbalance of ± 0.01 mg sensitivity (Mettler M3) in an environment with stable temperature and humidity, and the concentration was computed by dividing the weight gain of the filter by the volume of air sampled (Pfeiffer, 2005).
2.4 Chemical elemental determination in PM 10 and PM 2.5
The inductively coupled plasma mass spectrometer (ICP-MS) technique determined the samples’ elemental chemical composition. Filter samples were digested in a PTFE flask with 3 mL of HNO3 (suprapure grade, Merck, Darmstadt) at 120 ºC for 4 h on a hot plate. Then, the digested samples were diluted to 5% HNO3. The solution obtained was analyzed using an Elan DRC II ICP-MS (PerkinElmer SCIEX, USA). To correct instrumental drifts and plasma fluctuations, 103 Rh was used as an internal standard. Certified multi-element standards of trace elements were used to construct a six-point external calibration to quantify trace elements (Huamán et al., 2019). The following 33 trace elements were measured in the samples: Al, Sb, As, Ba, Be, Bi, B, Cd, Ca, Co, Cu, Cr, Sn, Sr, P, Fe, Li, Mg, Mn, Hg, Mo, Ni, Ag, Pb, K, Se, Si; Na, Ta, Ti, V, and Zn. The precision of the ICP-MS analysis was about 6% in the elements detected.
2.5 Data analysis
Descriptive statistics was used for concentrations of trace elements contained in aerosols, particle counting, and BC concentration for the entire campaign to characterize temporal behavior. Concentrations of trace elements of PM2.5 and PM10 samples were evaluated with Spearman’s multiple correlation tests (p < 0.05) to establish statistically significant dependencies.
The enrichment factor (EF) was calculated to study the influence of anthropogenic or natural activities on the measured chemical element. Titanium was used as a crustal reference element for the EF, following the studies of Bazzano et al. (2015) and Marina-Montes et al. (2020). Calculation of the EF was performed according to Eq. (1) (Rahn, 1971):
where X is the concentration of each trace element, X/T i air is the concentration ratio between element X and the reference T i element in the aerosol sample (PM2.5 and PM10), and X/T i crustal is the concentration ratio between the element X and the reference T i element in the soil surrounding the sampling site, given its abundance (in ppm) in the Earth’s crust (Wedepohl, 1995). Calculated EF values < 5 are explained as having a crustal or soil origin (Megido et al., 2017; Marina-Montes et al., 2020); if EF > 10 it indicates an enrichment attributable to additional sources, a significant fraction of anthropogenic origin (Chester et al., 2000; Bazzano et al., 2015; Marina-Montes et al., 2020).
Hierarchical cluster analysis (HCA) using Ward’s method of linkage and the squared Euclidean distance for similarities was used to segregate associations within the group of trace elements to identify possible pollutant sources (Hastie et al., 2009). All statistical analyses were performed using the CRAN R free software, v. 3.3.2 (R Team Core, 2015), using the following R packages: factoextra (Kassambara and Mundt, 2016), ggplot2 (Wickham, 2016), cluster (Chavent et al., 2012; Maechler et al., 2015), and psych (Revelle., 2018).
To analyze the influence of air masses from distant areas on trace element variability, the Hybrid Lagrangian Single-Particle Integrated Trajectory (HYSPLIT) model (Stein et al., 2015) from the National Oceanic and Atmospheric Administration (NOAA) Air Resources Laboratory (ARL) service was employed. The inverse trajectories (back-trajectories) were generated for 120 h (5 days backward) and 10 m above the ground. Meteorological reanalysis data from the Global Data Assimilation System (GDAS) with a horizontal resolution of 0.5º were used for running the model. Finally, the trajectories were grouped by chemical monitoring period and divided into eight different wind directions: north (N), northeast (NE), east (E), southeast (SE), south (S), southwest (SW), west (W), and northwest (NW).
3. Results and discussions
3.1 Chemical composition
Table I shows the average concentrations, standard deviation, and minimum and maximum values of the trace elements detected in PM2.5 and PM10 samples for the whole measurement period. For PM2.5, 12 elements were recorded, with the most abundant being, in decreasing order: Na > Si > Mg > Fe > Ca > Al; V (0.0012 µg m-3) and Cu (0.0013 µg m-3) were only detected once in the first and second sampling periods (February 7-9 and 10-13, respectively). In PM10, 16 trace elements were recorded, with the most abundant being, in decreasing order: Na > Fe > Si > Ca > Mg > Al. Other elements (As [0.0012 µg m-3], Hg [0.00015 µg m-3], Cu [0.0034 µg m-3], Pb [0.0021 µg m-3], and Zn [0.012 µg m-3]) where only detected once in the second sampling period.
PM2.5 (µg m-3) | PM10 (µg m-3) | ||||||
Parameters | Mean ± sd | Min-max | Parameter | Mean ± sd | Min-max | ||
PM2.5 | n = 5 | 4.04 ± 2.8 | 1.646-8.722 | PM10 | n = 5 | 12.54 ± 8.72 | 4.591-25.721 |
Al | n = 5 | 0.065 ± 0.07 | 0.00517-0.126 | Al | n = 5 | 0.141 ± 0.108 | 0.0172-0.280 |
Ba | nd | Ba | n = 2 | 0.00179 ± 0.000151 | 0.00169-0.00190 | ||
Ca | n = 5 | 0.068 ± 0.03 | 0.0363-0.127 | Ca | n = 3 | 0.340 ± 0.238 | 0.174-0.612 |
Fe | n = 5 | 0.097 ± 0.12 | 0.00823-0.291 | Fe | n = 4 | 0.454 ± 0.362 | 0.0463-0.921 |
K | n = 3 | 0.0127 ± 0.009 | 0.00641-0.0237 | K | nd | nd | |
Mg | n = 5 | 0.13 ± 0.09 | 0.0494-0.288 | Mg | n = 5 | 0.318 ± 0.17 | 0.135-0.545 |
Mn | n = 2 | 0.00187 ± 0.0009 | 0.00122-0.00251 | Mn | n = 3 | 0.00632 ± 0.004 | 0.00280-0.0107 |
Na | n = 5 | 0.933 ± 0.816 | 0.0352-2.24 | Na | n = 5 | 2.16 ± 0.854 | 1.17-3.24 |
P | n = 3 | 0.0174 ± 0.00884 | 0.00720-0.0230 | P | n = 5 | 0.0246 ± 0.0232 | 0.00445-0.0621 |
Si | n = 4 | 0.420 ± 0.23 | 0.208-0.815 | Si | n = 5 | 0.380 ± 0.07 | 0.291-0.455 |
Sr | Nd | Sr | n = 3 | 0.0009 ± 0.0004 | 0.000494-0.00122 | ||
Ti | n = 2 | 0.002 ± 0.0007 | 0.00148-0.00244 | Ti | n = 3 | 0.0046 ± 0.002 | 0.00240-0.00687 |
sd: standard deviation; n: number of samples where element was detected; min: minimum; max: maximum; nd: not detected
PM10 shows more significant variability compared to the PM2.5; likewise, the element with the greatest change in values in both particle sizes was Na. Considering the proportions of the elements detected in the PM2.5/PM10 index, Na registered 43%, Al 47%, Mg 41%, Fe 21%, and Ca 20% in PM10, which indicates a dominance of these elements in the coarse fraction. The abundance of Na in PM10 and PM2.5 coincides with results obtained by Artaxo et al. (1992) at the Brazilian Antarctic Station Comandante Ferraz, the nearest scientific station to our study site.
3.2 Enrichment factor
The enrichment factors determined during the chemical characterization sampling periods (Fig. 2) for PM2.5 show values greater than 10 (indicating severe enrichment) for Mg, Na, P, and V. The behavior was similar in PM10 for the first period, except for V, which was undetected. For the second period, the elements with EF values greater than 10 were Fe, Cu, and P in PM2.5 and Mg, P, Na, As, Ca, Cu, Fe, Hg, Mn, Pb, and Zn in PM10. In the third period, only Na had a value higher than 10 in both particle sizes. Similarly, for the fourth period, Mg, Na, and P in PM2.5 and Na in PM10 showed values greater than 10. In the last period, Mg and Na had values greater than 10 in both particle sizes, and P exceeded 10 in PM10.
Hierarchical cluster analysis (HCA) was applied to the concentration of trace elements, revealing the existence of three distinct groups in both particle sizes (Fig. 3). In the case of PM2.5, Group I was formed by K, Si, Ca, Mg, and Na of marine origin. In contrast, Group II was formed by Ti and Mn, which are indicators of ocean sediments (Aquilina et al., 2014). Group III was formed by Fe, Al, and P, which generally have the Earth’s crust as primary source. For PM10, Group I was formed by Mg, Na, P, Al, and Fe. Group II was formed by Sr, Ti, Ca, and Mn, and Group III was formed by Ba and Si, indicating that Group I has its origin in marine for dominance of sea-salt elements, Group II in earth crust, and Group III in continental weathering, thus it may be used as an indicator of primary productivity and meltwater input (Pyle et al., 2017). It should be noted that in both sizes, groupings of elements of marine and crustal origin have been identified as common sources, which aligns with the results of Artaxo et al. (1992).
3.3 Atmospheric particle size distribution
Increases in particles were observed on February 6, 10, 11, 16, 21, and 24, being more sustained from February 10 to 11, the days of the second chemical sampling period with the particle samplers (Fig. 4). Only in this period were recorded trace elements of anthropogenic origin such as As, Hg, Cu, Pb, and Zn. The colored shadow bands shown in Figure 4 represent the aerosol and chemical characterization sampling periods.
3.4 Black carbon
Increases in BC concentrations were observed on February 6, 9, 15, 21 and 24 (Fig. 5a), with the highest amount recorded on February 9, corresponding to the first chemical sampling period, with an apparent sky, where the highest concentrations of Ca, K, Mg, Mn, Na, P, and Si were observed, in addition to the only record of V in PM2.5, which would be related to activities linked to petroleum, together with fuel combustion processes at several research stations in the Antarctic subcontinent (John et al., 2023). On February 9 there were back trajectories of air masses that originated in the South American continent (Fig. 5b).
3.5 Trajectories
To investigate the possible origin of air masses contributing to the presence of aerosols over ECAMP during the studied period, a surface-level back trajectory analysis was conducted, separated by periods of chemical monitoring. The trajectories for each monitoring period were calculated with a 6-h interval and a backward duration of 120 h. Based on the obtained back trajectories (Fig. 6a), during the first monitoring period (February 7-9), the air masses originated from the South Pacific Ocean and the western side of the South American continent. In the second period (February 10-13), 25% of the air masses came from the South American continent, 50% from the Pacific Ocean, and the rest from the western Antarctic Ocean. In the third period (February 14-17), 73% of the air masses originated from the South American region, and 27% began in the South Pacific Ocean and crossed the southern tip of the South American continent before reaching the study site. In contrast, during the fourth period (February 17-20), 94% of the air masses came from the east, from the South Atlantic Ocean and western South Atlantic Ocean, with only 5% showing continental South American influence. For the fifth period (February 20-23), the majority (93%) of the air masses came from the southern Pacific Ocean and 7% from the Atlantic Ocean, indicating variability in air mass behavior.
When analyzing the frequencies of back trajectories divided into eight wind directions (Fig. 6b), the west (W) direction was prevalent during the first, second, and fifth periods, with frequencies exceeding 50%, followed by the northwest (NW) direction. During the third and fourth periods, the frequencies of the east (E), north (N), and northeast (NE) directions increased, particularly noticeable in the third period, where the N direction had its highest frequency, coinciding with a higher inflow of air masses from the South American continent (more than 70%). In this period, Na showed an EF more significant than 10, indicating that the predominant sources were maritime.
When relating wind direction frequencies to PM10 and PM2.5 concentrations determined by filter collectors, PM2.5 showed positive correlations with the southwest (SW) (r = 0.41; p > 0.05) and west (W) (r = 0.6; p > 0.05) directions and negative correlations with the N (r = -0.3; p > 0.05), E (r = -0.56; p > 0.05), NE (r = -0.88; p > 0.05), and south (S) (r = -0.45; p > 0.05) directions. For PM10, positive correlations were observed with the W (r = 0.8; p > 0.05) and NW (r = 0.3; p > 0.05) directions, and negative correlations with N (r = -0.4; p > 0.05), E (r = -0.88; p > 0.05), NE (r = -0.88; p > 0.05), and S (r = -0.34; p > 0.05) directions. This indicates that air masses from the W direction influence atmospheric particles, showing a stronger relationship with PM10. The W directions was more frequent in the first two periods of chemical sampling (Fig. 6c).
Correlating PM10 and PM2.5 concentrations with the most abundant elements in both particle sizes, Ca (r = 0.7; p > 0.05), Fe (r = 0.9; p < 0.05), Mg (r = 0.9; p < 0.05), and Na (r = 0.7; p < 0.05) showed the highest correlations with PM10. This suggests that the elements detected throughout the study period are primarily derived from coarse material, and local sources have more influence than remote ones. This information is supported by the results observed in the dendrograms, where they formed groups corresponding to sources of marine origin, crustal (soil), sedimentation, and rock weathering, as reported by Artaxo et al. (1992). In the study area, as in other places, the thawing of snow on glaciers has been occurring, releasing sedimentary rocks, gravels, and sands that, due to wind action, are carried and transported to other areas.
Figure 6c shows wind roses obtained from the meteorological station. During the first sampling monitoring period, the presence of V in PM2.5 was recorded, which coincided with an increase in the concentrations of BC, being an indicator of sources related to the burning of fuel from the research stations and maritime transport. On February 9 tourist cruise ships were observed in the area, in addition to trajectories of air masses originating in the south of the South American continent, which could indicate that there are remote sources of BC that can reach the study area.
In the second period, elements such as As, Hg, Cu, Pb, and Zn were detected in PM10 corresponding to an EF greater than 10, which indicates their anthropogenic origin; when examining wind speeds, the highest wind speed and highest PM10 concentration were recorded in the second period (Fig. 6d, suggesting that the elements with excessive enrichment were transported from nearby sources, being the ECAMP power generator (which operates with diesel fuel) the most likely source.
The results obtained in this study indicate that local sources, such as marine and crustal, are the main contributors to the studied atmospheric particles, which agrees with the findings published by Artaxo et al. (1992). An important aspect to consider is the sources of weathering, since with the retreat of glaciers the effect of wind erosion is an essential source in the area.
4. Conclusions
Antarctica can be influenced by the transport of atmospheric PM from nearby and distant locations. It is crucial to characterize the particles to identify their sources, especially in the Antarctic Peninsula, which is closer to urban places in South America.
We were able to characterize the daily variation of particle size distribution in eight size ranges, where it was possible to note peaks related to aerosol transport.
Measurements of black carbon concentration indicated the influence of local combustion related to fossil fuel use from the scientific station.
The chemical characterization of PM10 and PM2.5 showed a predominance of elements related to marine origin, as expected from the area’s characteristics. It also indicates the release of geological material that could contribute to the presence of elements such as Si, Ti, Al, and K in the PM.
There are external contributions (South America) and local contributions of black carbon such as burning, which, with the synergy of particles, could accelerate the melting of snow on the white continent.