Sampling sites

The urban site was located at the Centro de Ciencias de la Atmósfera (CCA) located at 19°19’34” N latitude and 99°10’55” W longitude inside the Universidad Nacional Autónoma de México, campus Ciudad Universitaria (CU), in the southern part of Mexico City at 2200 m a.s.l. The University has many green areas with high to moderate traffic densities. The other sampling site is a forested rural area (Tlalnepantla), State of Morelos, Mexico, located at 19°00’28” N latitude and 98°59´51” W longitude in the southern foothills of the Eje Neovolcánico at 2060 m a.s.l. Tlalnepantla borders on the north with the metropolitan area of Mexico City and the State of Mexico, on the southeast with Totolapan, the west with Tepoztlán, the south with Tlayacapan, and Cuautla City is further to the south and has moderate human activity. Pine and caducifolious forests lie to northeast and the northwest of the sampling site, with light vehicular traffic (Figure 1).

Figure 1 Localization of sampling sites at Centro de Ciencias de la Atmósfera, Universidad Nacional Autónoma de México, campus Ciudad Universitaria in Mexico City, and Tlalnepantla, State of Morelos, Mexico. 

A total of 145 rainwater samples were collected using automatic wet/dry precipitation collectors (Andersen, General Metal Works, Inc.) for inorganic ions on the weekends from mid-May to the end of October, which is the rainy season in central Mexico, during the 2006 to 2009 period. The samples were collected in standard high-density polyethylene (HDPE) buckets. To minimize contamination, the collector lid arms were Teflon-coated and the original aluminum cover and plastic-covered pad were replaced with a Teflon lid and pad. In the rural area Tlalnepantla, the sampling was done in passive collectors that consisted of a funnel attached to a high-density polyethylene (HDPE) bottle supported by a rod at a height of 1.5 m above ground level (^{Galloway et al., 1983}). After the rainy events, the collection buckets were tightly sealed with polyethylene covers and were sent to the chemical laboratory of the CCA. At the laboratory, all collected samples were filtered through 0.40 µm polycarbonate membrane filters, which were leached with 300 ml of deionized water. For the ion analysis, samples were poured into HDPE bottles, capped, and kept at 4 °C until analysis. The buckets used to collect the rain samples for major ion analysis and the flask and storage bottles were previously washed, brushed, and rinsed with distilled water and then deionized water (DI). The sample was treated under a clean hood (EPA) maintained at positive pressure to limit the introduction of contaminants from the laboratory.

Chemical analysis

The rainwater pH was measured by using an Orion 960 auto chemistry system within 24 h after the samples were collected. SO₄ ^2-, NO₃ ^- and Cl^- were analyzed by non-suppressed ion chromatography, and NH₄ ⁺ was measured by suppressed chromatography using a Perkin Elmer instrument equipped with an isocratic LC pump 250 and a ConductoMonitor III model conductivity detector. Na⁺, K⁺, Ca²⁺, and Mg²⁺ were analyzed by flame atomic absorption spectrometry with a GBC 932AA at 589, 766.5, 422.7 and 282.5 nm, respectively, with hollow cathode lamps (Photron Super lamp). For the background correction and analysis, a deuterium super lamp was used. The detection limits in µeq L^-1 were 4.58, 2.74, 1.13, 2.33, 0.074, 0.16, 0.5 and 0.13 for SO₄ ^2-, NO₃ ^-, Cl^-, NH₄ ⁺, Na⁺, K⁺, Ca²⁺ and Mg²⁺, respectively. The conductance was measured with an YSI 3200 conductivity instrument. The alkalinity was measured using the Grand`s titration method with an Orion 960 autochemistry system, and the HCO₃ ^- concentration was computed using the equation described by ^{Stumm and Morgan (1981)}. High-purity, traceable standards from the National Institute of Standards and Technology (NIST) were used to obtain the calibration curves for each of the major ions.

Quality control

The blanks of all glass and plastic were analyzed. The concentrations of blanks were below the major ion detection limits. The data quality for the collected samples was verified by comparison of the measured and calculated conductivity and by ionic balance evaluation, defined by ^{Peden et al. (1986)} as follows: Ion Percent Difference=[(Cations-Anions)/Cations+Anions)]×100 and Conductance Percent Difference=[(Calculated Conductance /Measured Conductance)]×100.

Statistics

Statistical analysis was performed with the integrated results of each analyzed ion. In this study, Statistica/W Stat Soft program (^{StatSoft, version 11, 2010}) was used for the data treatment and analysis and involved three steps: a) experimental data set arrangement and correlation efficient matrix building, b) data exploration and description by principal component analysis, and c) principal component factor loadings.

Correlation analysis

Spearman Rank Order correlation coefficients were calculated for the total ions. Table 2 show the matrix correlations for the different ions. If the coefficient between ions is high, then the source of emission is the same for both. Strong correlations were found between SO₄ ^2-, Cl^-, NO₃ ^-, Ca²⁺, Mg²⁺, Na⁺, K⁺ and NH₄ ⁺ from Morelos and CU; these ions were assumed to be correlated because they are completely soluble in rainwater. The strong correlation between Ca²⁺, Mg²⁺, Na⁺, and K⁺ also indicates a contribution from crustal material.

Factor analysis

To assess the relationships between the concentrations of the studied elements, a Principal Component Analysis was applied to normalize the varimax rotation. The varimax rotation aimed to maximize the variances of the squared normalization factor loadings across the variables for each factor and to facilitate interpretation. This method is the most widely used (^{StatSoft, version 11, 2010}). For the CU data, the principal component analysis was applied (Table 3), and four factors that explained 84.7 % of the total variance were chosen. In factor 1 (38.2 %), NH₄ ⁺ and SO₄ ^2- are associated with high loading and can be related to anthropogenic sources. In factor 2 (22.2 %), there is an association with NO₃ ^-. The two ions with highest loadings are Ca²⁺ and H⁺, suggesting that Ca²⁺ has a crustal origin, while H⁺ shows an association with the atmospheric acids (sulfuric and nitric), and NO₃ ^- from anthropogenic sources. In factor 3, Cl^- and Ca²⁺ are associated with high loadings with a moderate loading for SO₄ ^2-; these results indicate anthropogenic origins for Cl^- and SO₄ ^2- and a crustal origin for Ca²⁺. From the principal component analysis, four factors were chosen to explain 66.9% of the total variance of all of the data from Morelos (Table 3). The first factor (35% of the total variance), suggests no association of the species NO₃ ^-, SO₄ ^2- , and NH₄ ⁺ with anthropogenic sources and that Ca²⁺ is mainly of crustal origin. The second factor (13.2%) highlights the associations of Cl^-, Mg²⁺ and Na⁺, although these associations suggest marine origins. In the cases of the Mexico City and Morelos sampling sites, a marine origin is ruled out because both sampling sites are in a valley located at approximately 450 km from both the Pacific Ocean and Gulf of Mexico; therefore, the association of Cl^-, Na⁺ and Mg²⁺ in Morelos suggests an origin from soil dust and industrial emissions. Factor 3, with a variance of 10.4%, shows the association of NH₄ ⁺, NO₃ ^- and SO₄ ^2- with a high loading, which indicates anthropogenic sources, while the moderate loadings of Ca²⁺ and K⁺ point to crustal origins.

Acidifying and neutralizing components

Significant correlations between the sum of acidifying ions (SO₄ ^2-+NO₃ ^-+Cl^-) and the sum of neutralizing components (Ca²⁺+NH₄ ⁺+Mg²⁺+K⁺) were observed at CU (Figure 3a) and Morelos (Figure 3b). The pH values from CU were higher than the values from Morelos and varied between 5.67 to 6.11, which apparently indicate that the acidity of the precipitation was neutralized by the alkaline particles and NH₄ ⁺; because the pH values in Morelos were acidic (4.28 to 5.27), neutralization most likely did not occur because of the forested characteristics of the sampling site. It is known that the acidity of rain is mainly determined by strong acids, e.g., H₂SO₄ and HNO₃ (^{Overrein et al., 1980}. SO₄ ^2- y NO₃ ^- are the main acidifying components of rain and enable the fractional acidity (FA) to be calculated: FA=H⁺/(SO₄ ^2-+NO₃ ^-) (^{Balasubramanian et al., 2001}). If the ratio is equal to one, the acidity generated by these strong acids is likely not to be neutralized at all. Figures 4a and 4b show the variation in FA in the CU and Morelos rainwater from 2006 to 2009. In Mexico City from 2006 to 2009, a significant increase in FA (r = 0.47) was observed, indicating a small decrease in acidification. In Morelos, a significant decrease occurred, which indicates an increase in acidity. The relative contribution of NO₃ ^- to the acidification was calculated using the NO₃ ^-/(NO₃ ^-+SO₄ ^2-) ratio. In CU, the four-year average ratio was 0.48, indicating that 48 % of the rain acidity was due to NO₃ ^- and 52 % was due to SO₄ ^2-. In Morelos, the average ratio was 0.26, 26 % of the acidity was due to NO₃ ^-, and 74 % was due to SO₄ ^2-. Table 1a and Table 1b show that NO₃ ^- was the third most abundant ion in CU and Morelos after NH₄ ⁺ and SO₄ ^2-, representing an important contribution to the rain acidity in CU and even more so in Morelos.

Figure 3 Correlations between the sum of acidifying and that of neutralizing components for data of rainwater collected during the rainy seasons of years 2006-2009 in (a) Ciudad Univeristaria (CU) and (b) Morelos, sampling site. 

Figure 4 Variation of Fractional Acidity (FA) for data of rainwater samples collected during the rainy seasons of years 2006-2009 in (a) Ciudad Univeristaria (CU) and (b) Morelos.  

Some authors have associated the pH of rain with the dissolution of ions contained in aerosols. The pH of CU samples were higher than the values of Morelos samples and varied between 5.43 to 5.0 in the four years reported; indicating a slightly acid atmosphere, influenced by SO₄ ^2-, Cl^- and NO₃ ^- in 86 %. The rainwater acidity could make some ions highly soluble. The ion NH⁴⁺ is the most important neutralizing species, though the concentrations of Ca²⁺ + Mg²⁺ + K⁺ are additional contributions to the atmospheric neutralizing process (Table 4) (^{Ceron et al., 2005}; ^{Zhang et al., 2012}: ^{Cerqueira et al., 2014}, ^{Zhang et al., 2018}).

Back trajectories

There are many influences on precipitation chemistry. The transport paths of air masses before rainfall provide useful knowledge to evaluate and understand the chemical analysis of precipitation. Therefore, it is important to determine the variations in rainwater chemical composition by back trajectory analysis to estimate the sources of the emissions. The complex topography of the Mexico City valley and the surrounding areas could produce inconsistencies when coupling the rainwater ionic concentrations to the air mass back trajectory analysis. Therefore, a detailed analysis of the wind flow patterns based on reliable soundings at various altitudes above ground level in different zones of the Mexico City Valley is needed to obtain an accurate, high-resolution air mass back trajectory analysis. Due to journal space, only two trajectories per year are shown in Figure 5 for the CU and Morelos sampling sites. According to the back trajectory analysis (Figure 5) the prevailing wind directions from where the strongest wind originates are from W-SW, NW-W, and NE-N.

Figure 5 Backward modeled trajectories corresponding to the 2006-2009 period to each site: Ciudad Universitaria (CU) and Morelos. The blue squares and line indicate the modeled trajectories for a higher atmosphere circulation and the red triangles and line indicate the modeled trajectories for a lower atmosphere circulation for both sites (CU and Morelos). The vertical scale on the lower plot are meters above ground level, and the horizontal scale indicates the date