2.1 Meteorology

PEMBU provided data on wind speed and direction, temperature, solar radiation, pressure, and humidity. The Red Universitaria de Observatorios Atmosféricos (University Network of Atmospheric Observatories, RUOA) provided the greenhouse gas concentration data, and RAMA the air criteria pollutants concentrations.

HYSPLIT (Hybrid Single-Particle Lagrangian Integrated Trajectory) is a trajectory model that calculates simple trajectories of air parcels and simulates the transport and dispersion of particles in the atmosphere. We used it to calculate the backward trajectories of parcels and determine the possible origin of air masses impacting the site to establish source-receiver relationships. We used HYSPLIT to calculate the backward trajectories at 50 m above the ground level (AGL) with 0.5-degree global meteorological information (GDAS) and 24 hours backward.

2.2 Air quality

The RAMA routinely measures the concentration of criteria pollutants: NO, NO₂, CO, SO₂, O₃, PM₁₀, and PM_2.5, and reported one-hour average criteria pollutant concentrations. We downloaded the data from January 18 to March 31, 2015.

2.3 Particle sampling

We employed 47 mm Teflon and quartz filters to collect PM_2.5. Before sampling, Teflon filters were rinsed with deionized water and weighted on a semi-microbalance (Sartorious CPA225D). At the same time, quartz filters were preheated for six hours at 550 °C to eliminate all carbonaceous residues. They were conditioned for 24 h at 25 °C and constant relative humidity, and after that, weighed and stored individually in Petri dishes. After sampling, filters were placed in Petri dishes, wrapped with parafilm, and stored at 4 °C until the gravimetric analysis.

We collected 60 samples of PM_2.5 on quartz and 60 samples on Teflon substrates, both using MiniVol TAS (Airmetrics, USA) with a flow rate of 5 l min^-1 ± 10%, on 24 hours sampling basis.

We also collected 30 samples on quartz substrates (Pallflex 2500 20 × 25 cm, QAT-UP; Pall Sciences, Ann Arbor, MI, USA) using a high-volume sampler (Graseby Andersen SA-2000H) operating at 1.9 m³ min- for 48 hours. Samples were prepared and encapsulated in a tin crucible substrate for radiocarbon analysis.

2.4 Carbonaceous material

The carbon analyzer CM5014 (UIC, Inc) quantified the organic carbon (OC) and the total carbon (TC) of particle samples. The instrument measures the CO₂ gas stream produced from the complete combustion of carbonaceous material, which enters the coulombimetric cell and converts it into carbon content. The sample is heated at a fixed temperature. The OC evolved at 450 ºC and the TC at 700 °C (^{Alvarez-Ospina et al., 2016}). The difference between TC and OC corresponds to elemental carbon (EC), following the equation:

2.5 Elements

PM_2.5 collected onto Teflon substrates was employed for different analytical techniques to determine the chemical composition of particulate matter. X-ray fluorescence (XRF) analyses were performed with a spectrometer based on Oxford Instruments, an X-ray tube with Rh anode, and an Amptek Si-PIN X-Ray detector (resolution 160 eV at 5.9 keV) (Díaz et al., 2014; Reynoso-Cruces et al., 2021). The detector efficiency was measured using a thin film standard set (MicroMatter, Vancouver, Canada). The X-ray tube operated at 50 kV and 500 µA, irradiating each sample for 900 s. The analyses provided results on Al, Si, Fe, P, S, Cl, Cr, Mn, and Pb concentrations. Uncertainties were evaluated as described by (^{Hernández-López et al., 2020}).

Mercury (Hg) was analyzed using an atomic absorption spectrophotometer coupled to the hydride vapor generator (AAS-HVG) (Model GBC 932, GBC Scientific Equipment, Inc., Hampshire, Illinois, USA). Supra-pure HCl and HNO₃ digested the Hg according to Method IO-5, using an SEM microwave oven for elemental composition analysis in an OES-ICP (Thermo-Jarrel).

2.6 Water-soluble inorganic ions

A Dionex ICS-1500 ion chromatographer (IC) measured anions (NO₃ ^-, SO₄ ^-2, and Cl^-) and cations (Na⁺, K⁺, NH₄ ⁺, and Ca²⁺). Anions analysis was performed using a Dionex IonPac AS23 column (4 × 250 mm) and cations with a Dionex IonPac CS12A column (4 × 250 mm). The mobile phase was a carbonate solution (Na₂CO₃ 4.5 mM - NaHCO₃ 0.8 mM) for anions and methan- sulfonic acid (CH₄O₃S 20 mM) for cations at a flow rate of 1 ml min^-1 for both ions. The limit of quantification (LOQ) and determination (LOD) were calculated using the linear regression of standards calibration. LOD and LOQ were calculated as follows and are shown in Table 1.

Where σ is the standard deviation of the response and s is the slope of the calibration curve.

Also, an Aerosol Chemical Speciation Monitor (ACSM) measured the chemical composition of atmospheric aerosols in real time. It operates based on the principles of aerosol mass spectrometry, providing information on the concentrations of organic compounds and inorganic ions. Other authors have more information about the operation and calibration of ACSM during the measuring campaign (^{Prieto et al., 2023}; ^{Salcedo et al., 2018}).

We calculated the sulfur (SOR) and nitrogen (NOR) oxidation ratios to identify the oxidation rate of the atmosphere by using equations 4 and 5 (chemical species concentrations are in mol m^-3):

SOR and NOR estimate secondary transformation processes from NO₂ to NO₃ ^- and SO₂ to SO₄ ^2-. Values above 0.1 suggest photo-oxidation in the atmosphere might have occurred (^{Zhang et al., 2013}).

The neutralization ratio (NR) in the atmosphere refers to the balance between the most abundant acidic and basic components, often expressed as the ratio of acidic to basic ions or compounds in the atmosphere. It is related to acid-base reactions involving aerosols and gases. A neutralization ratio higher than one indicates an excess of basic ions relative to acidic ions, suggesting a more neutral or alkaline atmosphere. Conversely, a ratio less than one indicates an excess of acidic ions relative to basic ions, suggesting a more acidic atmosphere. It can help to assess the effectiveness of pollution control measures and better predict future changes in atmospheric composition. The NR is in equivalent concentrations:

2.7 Quality control and quality assurance

A tapered element oscillating microbalance (TEOM) was used as a reference for gravimetric analysis of PM_2.5. The instrument belongs to RAMA. The carbon analyzer was calibrated using the NIST SRM 1649a Urban Dust standard. For radiocarbon analysis by AMS, Oxalic acid II was used as a standard reference, and blanks with no radiocarbon [phthalic acid (C₈H₆O₄)] were also measured to subtract the background (Solís et al., 2017). The quality control and quality assurance (QC/QA) for Hg analyses were established using standard reference materials (SRMs), blanks, and blind samples in triplicate. The system calibration employed SRM NIST-2711 (Montana soil, certified Hg value 6250 ng g^-1), SARM-20 (Sasolburg bituminous coal, with Hg value of 250 ng g^-1), MESS-2 (Beaufort Sea estuarine sediment, certified Hg value 89 ng g^-1) and an internal laboratory standard (Hot Spring deposit, Hg value 330 ng g^-1). Accuracy verification for XRF was carried out using the NIST SRM 2783 (air particulate in filter media).

The curves with the standards were prepared with a multi-elemental standard (High Purity QC-TMFM-A) in 2% HNO₃ to QC/QA for elemental analyses, and duplicated analyses were carried out every five filters. SO₄ ^-2 (chromatographic method) and S (XRF analysis) showed a good correlation (r = 0.90). All chemicals were analytical grade, and solutions were prepared using ultrapure water (18.2 MΩ).

2.8 Statistical analysis

We averaged the concentration data from instrument measurements over 24 hours to match particle sampling times. We also calculated the basic statistics and kurtosis of the ion chromatographer and the ACSM to compare the data results between both instruments. We used the Matlab 2020 program (Mathworks, Inc.) to perform the statistical calculations and create the figures.

3.1 Meteorology

From January to March 2015, the predominant wind speed was below 1 m s^-1, blowing from west, south, and east. The temperature oscillated between 10 - 15 ºC, the relative humidity between 30 - 60%, and the rain was present only for two days. The atmospheric pressure oscillated between 773 and 780 mbar. All those meteorological measurements were common in the dry-cold season (winter) in the MCMA (Figure 2). The maxima daily solar radiation was above 600 W m², except on rainy days (February 14 and March 15).

Figure 2 Meteorological parameters ACU15, Left: a) average temperature, b) relative humidity, c) atmospheric pressure, and d) solar radiation. Right: wind rose. Data source: PEMBU. 

During the rainy days, the temperature diminished, and the relative humidity increased without significant changes in the wind speed. Based on the HYSPLIT backward trajectories and the wind rose of Figure 2, weak air masses traveled from the south at 50 m AGL toward the sampling site. Figure 3 shows three consecutive days in January, three days in February, and three more days in March 2015, with similar trajectories.

Figure 3 HYSPLIT backward trajectories January, February, and March 2015. 

3.2 Air quality

The average concentrations of O₃, NO₂, CO, and SO₂ were 27 ± 9 ppb, 23 ± 6 ppb, 0.79 ± 0.21 ppm, and 3.8 ± 3 ppb, respectively. NO₂, O₃, and SO₂ did not exceed the national standard air quality threshold, and CO was within the limit recommended by the World Health Organization (WHO). Figure 4 shows the time series of daily average concentration for criteria pollutants. O₃, NO, and NO₂ decreased on February 3 and 14 and March 15. In fact, on March 15, all criteria gases were at minimum concentrations because the rain scavenged gases and particles from the atmosphere.

Figure 4 Daily average criteria pollutants during ACU15. Data source: RAMA. 

The bottom chart in Figure 4 shows the time series of PM_2.5 reported by RAMA and sampled with impactors in ACU15. They have a correlation coefficient of 0.73. January 20 was windy, and February 14 and March 15 showed different concentrations. Those were days with rain.

3.3 Particulate matter

The maximum concentration threshold for PM_2.5 is 25 µg m^-3 hourly average (NOM-025-SSA1-2020), and sometimes the PM_2.5 concentration was above that number. The Secretaría de Medio Ambiente de la Ciudad de México (Mexico City Environmental Secretary, SEDEMA) reported a PM_2.5 annual average concentration of 21.3 µg m^-3 and an average daily maximum concentration of 40 µg m^-3 in 2015 (INECC, 2016). So, the average concentration of PM_2.5 (20.5 ± 8.0 µg m^-3) at the measuring site was lower than the annual average of that report, and the PM_2.5 concentrations did not show any significant differences between weekdays and weekends.

3.4 Carbonaceous material and radiocarbon content

Organic carbon (OC) was the most abundant species in PM_2.5, with a concentration almost three times higher than EC, corresponding to 75% and EC to 25% of TC contents. The average concentration of OC was 8.06 ± 1.85 µg m^-3, and for EC was 2.45 ± 1.69 µg m^-3. The OC/EC ratio ranged from 1.40 to 9.56, with a mean of 3.74 ± 1.79 (Table 4). Based on the weak intensity of the wind, vehicles and other local activities near the site were likely the primary sources of carbonaceous material. Furthermore, the predominant wind came from the south, where there are no industries, so the transport of pollutants probably came from vehicles. However, the OC fraction had significant variations. So, both primary sources and secondary processes contributed to the OC concentration, resulting in a poor correlation coefficient between OC and EC (r = 0.10). EC corresponded to 12% and OC to 39% of the total particle mass, meaning carbonaceous material was around 51% of PM_2.5.

The OC represents primary (POC) and secondary organic carbon (SOC) (^{Chow et al., 2002}; ^{Sharma et al., 2018}; ^{Turpin & Huntzicker, 1995}; ^{Vega et al., 2004}). SOC concentrations reported by other authors during ACU15 campaign had a poor correlation with the main components of secondary inorganic aerosols (SIA) (^{Rosa et al., 2024}). Table 2 shows the results of the cross-correlation analysis.

The radiocarbon analysis provides a tool for measuring fossil versus no fossil sources. The fraction of modern carbon (pMC) in fossil fuel combustion is zero, while no fossil sources have a pMC close to 100%. The pMC in PM_2.5 varied from 41% to 59%, with an average of 50% (n = 15), indicating biomass burning and other processes than fossil fuel combustion originated half of TC. If the total carbon from radiocarbon analysis (TCrc) has equal parts of pMC and fossil carbon (FC), then:

^{Rosa et al., (2024)} mention that the TC from thermal analysis (TCta) in PM_2.5 in the ACU15 campaign is formed by 25% primary organic carbon (POC), 50% secondary organic carbon (SOC), and 25% EC, so:

Equating equations 7 and 8 and assuming that EC and POC are exclusively fossil fuel combustion products, FC is formed by 0.5POC + 0.5EC, and the pMC corresponds to the SOC.

3.5 Elements

Al, Fe, and Si were grouped as soil elements related to the resuspension of road dust and earth crust material. Figure 5a shows a time series of soil element concentrations. Al and Si showed a correlation coefficient of 0.75, indicating the presence of aluminum silicates from soils (Figure 5b). The remaining elements (P, S, Cl, Cr, Mn, and Pb) were No soil elements (Figure 5c). Figure 5d shows P and S had a correlation coefficient of 0.91, meaning a solid association between sulfates and phosphates probably originated from gasoline combustion since both S and P are common additive agents in gasoline formulation (^{Hernández-López et al., 2020}).

Figure 5 Elements on PM_2.5, a) time series of soil elements, b) scatter plot Al vs. Si, c) time series of No soil elements, and d) scatter plot P vs S. 

Mercury was also present with an average concentration of 126 ± 3 pg m^-3; previous studies reported concentrations of 223 pg m^-3 and 187 ± 3 pg m^-3 in other sites of MCMA (^{Garza-Galindo et al., 2019}; ^{Morton-Bermea, Hernández-Álvarez, et al., 2021}; ^{Morton-Bermea, Schiavo, et al., 2021}; ^{Schiavo et al., 2022}). The sources of Hg are linked to industrial activities, such as pigment or preservative material in electrical and medical equipment, luminaries, and lubricating oils. The maximum concentration of Hg measured was 310 pg m^-3, which is higher than the mercury vapor limit established by the World Health Organization; those values were measured in two samples.

3.6 Secondary inorganic aerosol

The IC and the ACSM measured NH₄ ⁺, NO₃ ^-, and SO₄ ^2-, the most abundant components of SIA. In general, the IC measured lower concentrations than the ACSM. Table 2 shows the main results of both instruments.

According to ^{Salcedo et al., (2018)}, ACSM measures non-refractory material with 1.0 µm aerodynamic diameter or less and may have an error ranging from -30 to +10%, which probably explains the differences between both analytical instruments’ readings. However, the UIC and the ACSM measured similar concentrations of OC (Table 2).

The average total anion and cation concentrations measured by IC were 2.97 µg m^-3 and 1.31 µg m^-3, respectively. SO₄ ^2-, NH₄ ⁺, and NO₃ ^- correlated well with each other (Table 3), suggesting the presence of (NH₄)₂SO₄ and NH₄NO₃ produced by photochemical reactions due to emissions of gasoline and diesel exhaust compounds. The time series shows the equivalent concentrations of major (Figure 6a) and minor ions (Figure 6b) in equivalents.

Figure 6 Time series of ion composition in PM_2.5, a) major ions, b) minor ion, and c) sum of equivalents. 

The ratio K⁺/Cl^- was 0.88 ± 0.61, meaning that the source was not just the biomass burning. The ratio NO₃ ^-/SO₄ ^2- was 0.76 ± 0.63, suggesting that vehicles with fossil fuel engines are the primary source of those ions. There is a good neutralization of ions in the particle, so the NR is 1.06. Figure 6c shows the time series of the sum of cations and anions in equivalent concentration. The linear regression slope was 0.96, with a determination coefficient of 0.85, so particulate matter had a scarcity of anions. There are other particle anions (i.e., CO₃ ^2-, PO₄ ^3-, OH^-) that were not measured and probably neutralized.

SOC had correlations close to zero with SIA components, so particles incorporated diverse oxidized compounds, semi and low-volatile organics, biological material, and other sources of organic species that probably did not directly originate from atmospheric secondary reactions. Table 4 shows the average ratios of chemical species concentration.

PM_2.5, OC, and EC concentrations are like two other extensive previous studies (^{Chow et al., 2002}; ^{Vega et al., 2004}). However, the abundance of some ions has changed. Na⁺, K⁺, SO₄ ^2-, and NH₄ ⁺ are found in smaller amounts than previous studies, while Cl^- has increased. In the same way, Al, Mn, and Fe have increased. Crustal material and modifications in catalysts might cause those changes.

SO₄ ^-2 and NH₄ ⁺ showed a similar pattern between each other. Sulfate increased when the wind blew from the north. The average SOR was 0.35, indicating photo-oxidation of SO₂. The average NOR was 0.02. The average NO₃ ^-/SO₄ ^2- ratio was 1.12, which is common in areas where vehicular emissions dominate over coal or sulfur emissions, like the MCMA.

Table 4 shows the NOR, SOR, NR, OC/EC, and SOC statistics. SOR average is 0.35 and NOR 0.02. Both values are like those reported in 2021 (^{Millán-Vázquez et al., 2023}) for La Merced, a measuring site in Mexico City downtown. Both sites have identical oxidation rates, although Merced has a higher concentration of sulfates. The study of 2002 (^{Vega et al., 2004}) reported a pMC of 60% for an industrial zone and a pMC of 76% for a residential zone in Mexico City. The pMC estimated in ACU15 indicates a significant intensification (10% to 16%) in the contribution of fossil sources relative to previous measurements. That result is consistent with the substantial increase in the vehicular fleet in recent years and probably the loss of green areas in MCMA (^{Lima et al., 2023}).

We found a particle chemical composition like IMADA and 2000 studies. In 2021, ^{Millán-Vázquez et al., (2023)} found similar results in Merced in downtown Mexico City. For the previous studies, there have been no significant changes in carbon content, inorganic ions composition, or elements content in PM_2.5, probably meaning that the efforts in reducing and controlling the emissions sources of particles into the atmosphere have shown poor results (Table 5). For the study of 2021, the concentration of pollutants in ACU15 is lower than in Merced. Poor transport of contaminants from the south and minor local activities are probably the leading causes of low concentrations in ACU15.

Figure 7 shows the comparison of PM_2.5 composition with two previous studies. The concentration of PM_2.5, OC, EC, water-soluble inorganic ions, and elements is slightly lower in ACU15. The poor or null transport of material from other sites in the south, the local activities, and the diversity of particle sources probably explain the different particle compositions.

Figure 7 a) Mass composition and b) fraction of mass composition of PM_2.5 in Mexico City during three measuring campaigns (IMADA 1997, Vega et al., 2002, and ACU15 2015).