2.1 Meteorology

The Red Universitaria de Observatorios Atmosféricos (University Network of Atmospheric Observatories, RUOA) and the Programa de Estaciones Metereológicas del Bachillerato Universitario (University Program of High School Meteorological Stations, PEMBU) provided the meteorological data (wind speed and direction, temperature, solar radiation, pressure, and relative humidity) for the campaign. PEMBU has Vantage Pro 2 weather monitoring stations (Davis Instruments). Data can be consulted at https://www.ruoa.unam.mx/pembu/index.php?page=historical_facts

2.2 Particle optical properties

The instruments were located on the roof of the ICAyCC building, and the sampling probe was 5 m above the roof and 10 m above the ground to avoid resuspended soil particles. We used a photoacoustic spectrometer (PAS) and a photoacoustic extinctiometer (PAX) to measure the optical properties of PM_2.5. Both instruments operated continuously from January 21 to February 6 and from February 17 to March 21, 2015. We connected both instruments to a TSI aerosol splitter and then to a 2.5 µm cyclone (URG). PAS and PAX sampled aerosols with a flow rate of 1.0 l min^-1.

PAS and PAX have the same operation principle, where the quantity of absorption measured is proportional to the sound pressure produced in an acoustic resonator caused by the absorption of light. The acoustic resonator operates in plane wave mode with a resonant frequency of 500 Hz and a photoacoustic coefficient of 12.8 Pa (W m^-1). The coherent acoustic noise is suppressed with filters at the pressure nodes of the resonator. The laser wavelengths selected for the photoacoustic spectrometers avoid absorption of gaseous atmospheric light, with a gaseous absorption coefficient of 0.1 Mm^-1 (^{Liñán-Abanto et al., 2021}). The instruments have also a wide-angle integrating reciprocal nephelometer to measure the scattering coefficient of light.

The lower detection limit for light absorption is 0.4 Mm^-1, which corresponds to a mass density of elemental carbon of 40 ng m^-3, assuming a light absorption efficiency of 10 m² g^-1. Also, the absorption and scattering coefficients of atmospheric gases like O₂, NO₂, O₃, and H₂O are two orders of magnitude lower compared to those of aerosols (^{Horvath, 1993}).

The University of Nevada built the PAS, which has a laser operating at 532 nm (^{Arnott et al., 1999}). The PAX is a commercial model (Droplet Measurement Technologies, USA) with a laser operating at 870 nm. Both spectrometers recorded the particle’s optical properties (absorption, scattering, and extinction coefficients) every minute. PAX also quantified BC using a default mass absorption efficiency (MAE) of 4.74 m² g^-1 (^{Prieto et al., 2023}). The intercomparison of PAS and PAX data had a good determination coefficient (R² = 0.97).

2.3 PM _2.5 sampling

Before sampling, 47 mm quartz filters (Tissue quartz 2500QAT-UO, 47 mm, PALL) were preheated at 600 ºC for 6 h to eliminate the organic material that could be present. Then, we conditioned the substrates for 24 h at constant relative humidity (less than 30%) and 25 ºC, weighed them on an analytical balance (Sartorius CPA225D-OCE), and boxed them individually in plastic Petri dishes. The plastic dishes were wrapped with parafilm tape and opened until the sampling time. Filters after sampling were stored at 4 ºC, stabilized for 24 h at the same conditions as before and weighed.

We calibrated the Minivol sampler (Airmetrics) flow rate with a Gillibrator at a constant 5.0 L min^-1. The impactor collected PM_2.5 on filters every 24 h. The Red Automática de Monitoreo Atmosférico (Automatic Atmospheric Monitoring Network, RAMA) of Mexico City provided continuous PM_2.5 concentrations, with a Thermo Andresen model FH62C14 continuous particulate monitor based on beta attenuation.

2.4 Carbon analysis

The carbon analyzer CM5014 (UIC) measures OC and total carbon (TC) content. The instrument quantifies CO₂ produced from the complete combustion of carbonaceous material in a coulombimetric cell and converts it into carbon content. We calibrated the instrument with NIST SRM 1649a Urban Dust (^{Álvarez-Ospina et al., 2016}). OC evolved at 450 ºC and TC at 700 ºC. The difference between TC and OC corresponds to elemental carbon (EC), following Eq. (1):

2.5 Optical properties and statistical analysis

We measured the particle’s absorption and scattering coefficients every minute and averaged the values for 24 h to match the optical property databases with those of OC and EC analyses. With the daily averages, we calculated the MAE₅₃₂ (532 nm) and MAE₈₇₀ (870 nm) according to Eq. (2), which represents the ratio of particle absorption coefficient and the concentration of light-absorbing chemical species, basically EC. We also calculated the single scattering albedo (SSA) with Eq. (3):

We applied a linear regression of b _abs as a function of EC, where the slope corresponds to the MAE (m² g^-1). To contrast the hypothesis of normality, we used the Kolmogorov-Smirnov statistical test (α = 0.05). The data had a non-parametric distribution; hence, we applied the Mann-Whitney statistical test to compare b _abs and EC.

3.1 Meteorology

January, February, and March are part of the cold-dry season in Mexico City, and cold air masses often blow from the S-SW to the city. Southerly winds are associated with air masses containing low temperatures and high relative humidity. These air masses are not polluted, unlike those coming from the north, where the industrial area of Mexico City is located.

From January to mid-March the mean wind speed was 1.8 m s^-1. Wind speed was slow, so we believe local activities produced most of the airborne particles. By March 15, the average wind speed increased to 3.0 m s^-1 (Fig. 1). In January and February, the W-NW wind direction prevailed and in the week of March 8 to 15, the wind changed its direction and blew from the south (Fig. 1). From January 19 to February 15, the average temperature was 13.9 ± 1.5 ºC and from March 8 it increased to 17.8 ± 1.6 ºC (Fig. 1). From January 19 to March 8 relative humidity was 50% and after March 8 it increased to 60% with southerly winds. Figure 1 shows the basic meteorological parameters (wind speed and direction, temperature, and relative humidity) recorded in the measurement campaign.

Fig. 1 Time series of meteorological parameters: wind speed, temperature, wind direction, and relative humidity. Red thin lines are the 24-h moving averages. 

3.2 Particle optical properties

At 532 nm, the average b_abs was 13.68 ± 6.89 Mm^-1, and at 870 nm it was 10.40 ± 5.33 Mm^-1. From January 18 to 31, it decreased as the wind intensity increased. In February, b_abs varied from one week to another, ranging from 7.25 to 21.92 Mm^-1 (532 nm) and from 5.10 to 17.44 Mm^-1 (870 nm). From March 8 to 15, it decreased as the wind speed went up (Fig. 2).

Fig. 2 Particle’s optical properties: absorption coefficient, scattering coefficient, single scattering albedo, and PM_2.5. Dotted blue lines correspond to the 532 nm wavelength and orange lines to 870 nm. 

The average b_sca was 71.00 ± 40.56 Mm^-1 at 532 nm and 21.37 ± 12.31 Mm^-1 at 870 nm. Between January 21 and February 6, the scattering decreased to half of its average value, with the wind blowing from N-NW, probably transporting pollutants from the industrial part of the city. From March 8-15, b_sca decreased associated with a cold front entering Mexico City (Fig. 2).

At both wavelengths, the SSA reached its minimum value at 07:00, when light absorption was at its maximum produced by the presence of dark particles. The SSA₅₃₂ maximum was 0.89, and the SSA₈₇₀ maximum was 0.77, around 13:00 LT, when the absorption coefficient dropped to its minimum value (^{Prieto et al., 2023}). The average SSA₅₃₂ was 0.83 ± 0with.35 Mm^-1 and the average SSA₈₇₀ was 0.66 ± 0.28 Mm^-1. The difference in values is probably due to more chemical species in the particle being sensitive to lower wavelengths.

Particles are covered with different materials, oxidated species, water, and other coating substances changing the particle optical properties and promoting the formation of secondary aerosols. Furthermore, SSA₅₃₂ values were close to 1.0 (Fig. 2), probably linked to secondary aerosols and other materials that scatter light (^{Eck et al., 1999}; ^{Cheng et al., 2011}; ^{Schumann, 2012}; ^{Tao et al., 2019}).

3.3 Carbon content

The average mass concentration of PM_2.5 was 18.45 ± 8.00 µg m^-3, while the average concentration of OC was 7.96 ± 3.61 µg m^-3, and for EC it was 2.44 ± 1.77 µg m^-3. So, EC corresponded to 12% and OC to 39% of the total PM_2.5 mass, which means that the carbonaceous material is 51% of the particle mass. Vehicles and commercial activities were probably the main sources of emission of carbonaceous material near the sampling site. The carbon composition had significant variations, perhaps due to the contribution of various sources and secondary atmospheric processes. The correlation coefficient between OC and EC was 0.10. The average OC/EC ratio ranged from 1.08 to 22.89, with a mean of 5.58 ± 1.08 (^{Turpin et al., 1990}; ^{Chow et al., 2011}).

The was 1.08, and primary organic carbon (POC) was calculated following Eq. (4). We used Eq. (5) to calculate secondary organic carbon (SOC) (^{Castro et al., 1999}). SOC average concentration was 5.32 ± 2.96 µg m^-3, meaning that the secondary species contribution was more significant than POC. SOC is linked to temperature, but the statistical analysis concludes that there is no significant difference between SOC concentrations and monthly temperature (^{Zhang et al., 2013}). In January, February, and March, the average temperature was 14.54, 14.94, and 15.88 ºC, respectively.

Figure 3 shows the daily variations of organic carbon (POC and SOC) and EC. Variations in POC and SOC concentrations are likely due to changes in emission sources, transport of airborne material, or secondary processes in the atmosphere.

Fig. 3 Carbon content in PM_2.5 samples. Primary organic carbon (POC) and secondary organic carbon (SOC) were calculated following Eqs. (7) and (8). (EC: elemental carbon.) 

3.4 Mass absorption efficiency (MAE)

MAE efficiency corresponds to the ratio of daily average absorption coefficient and EC concentration, following Eq. (2). The range of MAE₅₃₂ was 0.11-2.44 m² g^-1 with an average value of 0.59 ± 0.57 m² g^-1. The range of MAE₈₇₀ was 1.39-25.60 m² g^-1, with an average value of 6.78 ± 5.85 m² g^-1. The PAX uses a default MAE of 4.74 m² g^-1 that falls into the average MAE₈₇₀.

The EC concentration and the daily average b_abs changed from one day to another, probably caused by changes in the emission rates of local sources, transport, or atmospheric aging processes. Hence, MAE at both wavelengths had considerable variations, and the correlation coefficients of EC versus b_abs were 0.41 (at 532 nm) and 0.39 (at 870 nm).

We sorted the data by OC/EC ratios in four groups assuming an external mixing of the particle. We assumed EC was the only absorbing material in the particle at both wavelengths and different concentrations of scattering material modified the absorbing properties of EC, like a shadowing effect (^{Weingartner et al., 2003}). Then we calculated the linear regression for each group of data. Groups 1 and 5 corresponded to OC/EC ratios < 2.0, where OC and EC concentrations were similar. Groups 2 and 6 were for OC/EC ratios between 2.0 and 3.1. Groups 3 and 7 comprised OC/EC ratios between 3.1 and 8.0. Groups 4 and 8 were for particles with a notable predominance of organic compounds, with OC/EC ratios greater than 8.0. Table I shows the OC/EC ratio ranges, the population, the linear regression equation, and the determination coefficient for each group. Figure 4 shows the data and linear adjustment for each group. All regressions intersect the b_abs axis at EC = 0 because we assumed EC was the unique contribution to light-absorbing in the particle and OC modified the light-absorbing properties of EC but did not absorb light directly.

Fig. 4 Scatter plots of elemental carbon (EC) and light absorption (b_abs) grouped by different OC/EC ratios, where MAE corresponds to the slope of the regression lines, at 532 nm and at 870 nm. (OC: organic carbon.) 

The carbonaceous material, with a high contribution of OC and other species, changed the optical properties of aerosols, and the amount of primary and secondary organic compounds modified the aerosols’ absorbing light capacity. When the amount of OC was like EC in the particle (OC/EC ratio < 2.0), we had a MAE₅₃₂ of 19.75 m² g^-1, and MAE₈₇₀ was 15.26 m² g^-1. As the amount of OC increased MAE diminished, and if the OC concentration was eight times the EC mass (OC/EC ratio > 8.0), then the MAE₅₃₂ was equal to 2.67 m² g^-1, and the MAE₈₇₀ was 2.03 m² g^-1.