Molecular Markers in Ambient Air Associated with Biomass Burning in Morelos, M é xico

: Atmospheric particles with an aerodynamic diameter less than or equal to 2.5 micrometers (PM 2.5 ) were collected at two sites located in the urban area of the city of Cuernavaca (Morelos) during a season when a large number of forest ﬁres occurred. Three dicarboxylic acids (malonic, glutaric and succinic) and levoglucosan were analyzed by liquid chromatography coupled with mass spectrometry (ESI-Q-TOF) and soluble potassium (K + ) was analyzed by ion chromatography. The concentration of PM 2.5 increased on the days when the highest number of forest ﬁres occurred. A strong correlation was observed between levoglucosan and K + , conﬁrming the hypothesis that both are tracers of biomass burning (r = 0.57, p < 0.05). Levoglucosan (average 367.6 ng m − 3 , Site 2) was the most abundant compound, followed by succinic acid (average 101.7 ng m − 3 , Site 2), glutaric acid (average 63.2 ng m − 3 , Site 2), and malonic acid (average 46.9 ng m − 3 , Site 2), respectively. The ratio of C 3 / C 4 concentrations ranged from 0.5 to 1.2, with an average of 0.8, which suggests great photochemical activity in the Cuernavaca atmosphere. The ratio of K + / levoglucosan concentrations (0.44) indicates that open ﬁres are the main source of these tracers. The positive correlations between PM 2.5 and levoglucosan and succinic and malonic acids suggest that such compounds are contributing to secondary organic aerosol particle formation.


Introduction
Biomass burning is a major source of particulate matter in ambient air, with the particles consisting of hundreds of organic and inorganic compounds, causing significant impacts on regional to global air quality [1], human health [2] and climate [3]. The composition of the atmospheric particles depends on diverse factors, such as the original source, the amount of emission, the meteorological conditions and

Sampling of Airborne Particles
Sampling was done between February and August 2016. The particulate material was collected on fiberglass filters of 47 mm diameter and 2.0 µm pore size (Whatman, Darmstadt, Germany), which were conditioned in an oven at 550 °C for 30 min, and then transferred to a desiccator for 24 h. A lowvolume (5.0 L min −1 ) sampler (MiniVol™ TAS, Eugene, OR, USA) was used, equipped with a 2.5 µm impactor. Sampling was carried out for periods of 24 h (12:00 a.m.-12:00 p.m.). Gravimetric analysis was carried out under controlled room temperature and humidity (22 ± 3 °C, 40% ± 5%). Before weighing, the filters were stabilized for 24 h. Three blank filters per month were used as laboratory blanks. Each filter was weighed at least three times and the three readings had to agree within 5 micrograms to be accepted. The detection limit was calculated as 3 × SD of the mass change in the blank filters divided by the volume of the corresponding exposure time (24 h). The detection limit for the measurements was 0.92 mgm −3 . The mass determination of PM2.5 was performed by weighing the filters before and after the collection period with a 0.1 µg sensitivity microbalance (Citizen CX-220 ™, Parwanoo, India). The concentration of PM2.5 in the atmosphere in mgm −3 was obtained by dividing the mass by the volume of filtered air, adjusted to conditions of standard temperature and pressure (25 °C and 760 mm Hg).

Meteorological Parameters and Criteria Pollutants
Simultaneously, meteorological parameters were monitored at both sites (temperature, relative humidity, solar radiation, wind speed and wind direction). Additionally, only at Site 1, criteria pollutant data were obtained (O3, SO2, NOx and CO) ( Table 1).

Sampling of Airborne Particles
Sampling was done between February and August 2016. The particulate material was collected on fiberglass filters of 47 mm diameter and 2.0 µm pore size (Whatman, Darmstadt, Germany), which were conditioned in an oven at 550 • C for 30 min, and then transferred to a desiccator for 24 h. A low-volume (5.0 L min −1 ) sampler (MiniVol™ TAS, Eugene, OR, USA) was used, equipped with a 2.5 µm impactor. Sampling was carried out for periods of 24 h (12:00 a.m.-12:00 p.m.). Gravimetric analysis was carried out under controlled room temperature and humidity (22 ± 3 • C, 40% ± 5%). Before weighing, the filters were stabilized for 24 h. Three blank filters per month were used as laboratory blanks. Each filter was weighed at least three times and the three readings had to agree within 5 micrograms to be accepted. The detection limit was calculated as 3 × SD of the mass change in the blank filters divided by the volume of the corresponding exposure time (24 h). The detection limit for the measurements was 0.92 mgm −3 . The mass determination of PM 2.5 was performed by weighing the filters before and after the collection period with a 0.1 µg sensitivity microbalance (Citizen CX-220™, Parwanoo, India). The concentration of PM 2.5 in the atmosphere in mgm −3 was obtained by dividing the mass by the volume of filtered air, adjusted to conditions of standard temperature and pressure (25 • C and 760 mm Hg).

Meteorological Parameters and Criteria Pollutants
Simultaneously, meteorological parameters were monitored at both sites (temperature, relative humidity, solar radiation, wind speed and wind direction). Additionally, only at Site 1, criteria pollutant data were obtained (O 3 , SO 2 , NOx and CO) ( Table 1).

Extraction of Water-Soluble Organic Compounds
The sampled filters were placed in polypropylene tubes, and then 8.0 mL of Milli-Q water was added to each tube and extracted in an ultrasound bath (Branson 3210) for 1.0 h. The Erlenmeyer flasks in which the filters were introduced were fitted with cooled condensers with water at 10 • C; then the extracts were filtered through nylon membranes with a pore size 0.45 µm. Once the extracts were filtered, they were transferred to injection vials and stored at approximately 4 • C, until chromatographic analysis.

Chromatographic Analysis
The analysis of the WSOC was carried out by using Ultra-High-Performance Liquid Chromatography (UHPLC) in an Agilent 1290 Infinity LC System, coupled to a quadrupole-time-of-flight (Agilent Q-TOF 6545 mass spectrometer), equipped with an electrospray ionization source (ESI). The ESI-MS conditions were optimized by injection of single standard at a concentration of 100 ng mL −1 of four compounds (levoglucosan and glutaric, malonic and succinic acids), in both positive and negative ionization mode. No signal was recorded in positive mode. Highly abundant analytes signals were detected in negative mode. Table 2 shows optimal instrumental operating conditions. For the separation of compounds, a Zorbax Rapid Resolution High Definition SB-C18 column with a 2.1 mm internal diameter × 50 mm and a particle pore size of 1.8 µm was used. The mobile phase was a solution of NH 4 OH 13 mM in methanol:water (85:15), which was used in isocratic mode. The identification of the compounds was carried out by the retention time and the molecular ion (Table 3). Meanwhile the analysis of soluble potassium (K + ) was determined by ion chromatography (CI, Metrohm model 861 Advanced Compact with conductivity detector), without chemical suppression in a Metrosep C 2 _150 (Metrohm) column; the mobile phase was a solution of tartaric-dipicolinic acid (4.0:0.75 mM) at a flow rate of 1 mL min −1 . The injection volume was 50 µL.
Calibrations for all studied compounds were based on serial dilutions from a stock solution made by dissolving individual compounds in solid form. Each calibration graph was made with five concentration points. The concentration range for levoglucosan and carboxylic acids was between 50 and 5000 ng mL −1 , while for K + , this was between 50 and 1000 ng mL −1 .

Quality Control of the Analytical Methods
To exclude the presence of the compounds of interest in the materials and reagents used, laboratory targets were done once a week. The extraction efficiency was determined by enriching filters with a mixture of the compounds studied at a concentration of 100 ppb for levoglucosan, malonic acid, succinic acid and glutaric acid and 500 ppb for K + . The enriched filters were extracted in the same way as the samples; the recovery percentages ranged from 80% to 85%. Furthermore, from the linear regression data, the detection limit of the method (LD) and quantification (LC) were determined, as well as the correlation coefficient (R) according to Miller and Miller (2002) [27] (Table 4).

Wind Trajectories
To determine the behavior of winds in the study area, HYSPLIT4 model trajectories, obtained from the NOAA (National Oceanic and Atmospheric Administration), were used [28]. During the study period, two patterns were observed with respect to the direction of the winds. The first occurred between February and April; in these months, the winds came mainly from the west-southwest. During the second period, from May to July, they came mainly from the east-southeast ( Figure 2).

Concentrations of PM2.5
In total, 35 samples were taken at each site, and the results revealed that the average concentration of PM2.5 was 22 µg m −3 for Site 1 (CENTRO) and 16 µg m −3 for Site 2 (CIQ). The behavior of the PM2.5 concentration was evaluated by time series (Figure 3). The results indicate some similarity, which is probably associated with common sources influencing both sites. The highest concentration of PM2.5 was observed in February and between April and May, which is consistent with the forest fires that occurred in this period ( Figure 4). According to official data, 52 fires occurred in February, 50 in March, 59 in April and 21 in May. This last month (May) is when the rains start,

Concentrations of PM 2.5
In total, 35 samples were taken at each site, and the results revealed that the average concentration of PM 2.5 was 22 µg m −3 for Site 1 (CENTRO) and 16 µg m −3 for Site 2 (CIQ). The behavior of the PM 2.5 concentration was evaluated by time series (Figure 3). The results indicate some similarity, which is probably associated with common sources influencing both sites. The highest concentration of PM 2.5 was observed in February and between April and May, which is consistent with the forest fires that occurred in this period ( Figure 4). According to official data, 52 fires occurred in February, 50 in March, 59 in April and 21 in May. This last month (May) is when the rains start, which allowed for the natural abatement of the fires and therefore the decrease in the concentration of PM 2.5 [22].
Atmosphere 2020, 11, x FOR PEER REVIEW 7 of 16 which allowed for the natural abatement of the fires and therefore the decrease in the concentration of PM2.5 [22].     Table 5 summarizes the concentrations of levoglucosan and succinic, glutaric and malonic acids associated with PM2.5 at both sampled sites in Cuernavaca and their average values with standard deviations, as well as maximum and minimum values. Table 5. Mean concentration and range of the selected molecular markers during study period (ng m −3 ).

Centro
Average SD Minimum Maximum  Table 5 summarizes the concentrations of levoglucosan and succinic, glutaric and malonic acids associated with PM 2.5 at both sampled sites in Cuernavaca and their average values with standard deviations, as well as maximum and minimum values.

Levoglucosan
Levoglucosan was the compound with the highest concentration at both sites. At the CIQ site, the average concentration was slightly higher (367.6 ng m −3 , ±57.6), suggesting similar sources during the study period. Levoglucosan is the most abundant anhydrosugar reported in urban areas, due mainly to the burning of biomass [29]. Meanwhile, low concentrations found in summer could be attributed to the absence of residential heating in the warm season and better atmospheric conditions that favor atmospheric dispersion of pollutants. For its part, other minor sources such as emissions from fires for land clearance and/or accidental fires could be possible sources of levoglucosan in summer [30]. In this sense, it is important to highlight that, in the state of Morelos, about 182 forest fires occurred during the study period, which probably justifies the high concentrations of levoglucosan. Moreover, in some way, the presence of levoglucosan at this site probably indicates the transport of air masses from other regions of the state, specifically east and east-southeast, considering the behavior of wind trajectories, during the study period [28].

Dicarboxylic Acids
The concentration of some organic acids was also determined. At both sites, the most abundant compound was succinic acid (C 4 ), followed by glutaric acid (C 5 ) and malonic acid (C 3 ). Again, it is observed that the CIQ site presents slightly higher concentrations than the Centro site. When evaluating the behavior of the concentrations, it is observed that the maximum values were presented in the period (April-May) in which the greatest number of fires were registered, which is congruent with the observed behavior of the concentration of PM 2.5 and Levoglucosan.
The total average concentration of organic acids (C 3 , C 4 and C 5 ) was 178.1 ng m −3 at the Centro site and 211.8 ng m −3 at the CIQ site. The excess concentration of succinic acid over malonic acid indicates that emissions can come from burning biomass, fossil fuel combustion and vehicular emissions [31,32]. Kawamura and Ikushima (1993) indicate that malonic acid is partly produced from the incomplete combustion of fossil fuels and biomass burning, but is mainly due to photochemical oxidation of succinic acid in the atmosphere [15]. On the other hand, Kawamura et al. (1996a) proposed that succinic acid can also be generated via the photo-oxidation of unsaturated fatty acids from terrestrial higher plants and domestic cooking [33].
Moreover, Kawamura and Sakaguchi (1999) proposed that the C 3 /C 4 ratio is a good indicator to determine if the presence of dicarboxylic acids in the atmosphere corresponds to primary sources or oxidation processes [31]. Kawamura and Kaplan (1987) reported that lower C 3 /C 4 ratios (0.25-0.44) correspond mainly to vehicular emissions, while values above that correspond to oxidation processes in the atmosphere [16]. The C 3 /C 4 ratio obtained in this study, ranging from 0.5 to 1.2, with an average of 0.8, suggests that, in the urban area of Cuernavaca, photochemical processes regulate, to a large extent, the presence of the dicarboxylic acids studied.

Comparisons of the Observed Concentration for Levoglucosan and Dicarboxylic Acids with Other Studies
These comparisons should be analyzed with caution, since the results were obtained with different analytical methodologies and under different environmental conditions. The average concentration of levoglucosan was lower than those reported in samples collected in Elverum, Norway (605 ng m −3 ), during the winter of 2002, in a suburban atmosphere, and the 713.0 ng m −3 recorded at Tengchon, China, during the spring of 2004, in a rural site, suggesting that the main sources of levoglucosan in such places can be attributed to the combustion of firewood and agricultural and garden-waste burning [34,35]. However, the concentrations of levoglucosan found in the present study were significantly higher than those observed in the rural atmosphere in K-puszta, Hungary  Table 6). Regarding the observed differences in levoglucosan concentration, Liu et al. (2005) mentioned that it is also important to consider different emission fluxes for different biomass combustion types, such as fuel types, oven types, agricultural fires and forest fires, among others [39].
Likewise, comparisons of dicarboxylic acid concentrations observed in this study with other studies conducted in other urban areas showed that the average concentration of dicarboxylic acids  [44]. Such differences can be attributed to an increase in burning for residential heating in cold periods [19]. However, other studies reported that some of the DCAs tend to be higher during the warm season in urban areas, mainly due to photochemical activity [45]. In this sense, it is important to highlight that the period in which the present study was carried out was a warm season with the presence of many forest fire events (Table 6).

Water-Soluble Potassium (K + )
To evaluate the degree of association between water-soluble potassium and levoglucosan, both considered as indicators of biomass burning, a time-series analysis was performed with the observed concentrations during the study period. Figure 5 reveals two peaks on the days when the highest number of forest fires occurred in the state of Morelos (April-May 2016), confirming the hypothesis that both are indicators of biomass burning. This behavior is consistent with that observed by Zhang et al. (2010) in a study conducted in the urban areas of some cities in the Southwestern United States during winter, when there was an increase in biomass burning [46]. Likewise, Reche et al. (2012) observed similar behavior in the urban area of Barcelona, in biomass burning scenarios [38]. Locker (1988) established the K + /levoglucosan relationship to determine if the combustion processes correspond to ovens or fireplaces that use wood, or open fires. If the K + /levoglucosan ratio is <0.2, it may indicate the prevalence of domestic heating with wood, while a K + /levoglucosan ratio around 0.5 may indicate open fires or combustion of fuels such as slash, or straw [47]. In the present study, the value was 0.44, probably indicating that open fires are the main source of these tracers in the study region.

Sources of Molecular Markers
In order to investigate possible correlations between the studied compounds, meteorological data and criteria pollutants, the Spearman correlation coefficients were calculated (  Figure 5. Association between water-soluble potassium and levoglucosan.

Sources of Molecular Markers
In order to investigate possible correlations between the studied compounds, meteorological data and criteria pollutants, the Spearman correlation coefficients were calculated (Table 7).  Levoglucosan positively correlated with glutaric and malonic acids and K + , probably indicating common sources, i.e., from biomass burning and mainly cellulose pyrolysis [48,49]. The positive correlation observed between temperature and O 3 with glutaric and malonic acids and levoglucosan suggests the incidence of photochemical activity. Specifically, for the observed correlation between temperature and malonic and glutaric acids,  indicates that this behavior is associated with secondary photochemical processes in the atmosphere rather than primary emissions from vehicular exhaust [19]. Other studies suggest that glutaric acid has a mechanism of formation through secondary photochemical reactions in the atmosphere, from vehicle emissions [33,50]. The positive correlation observed between temperature and levoglucosan can be attributed to the large number of forest fires that occurred in the study season in the state of Morelos. The positive correlation between PM 2.5 and glutaric and malonic acids and levoglucosan suggests that such compounds are contributing to secondary organic aerosol particles' formation. Gaseous pollutants (NO x , CO and SO 2 ) did not show a correlation with levoglucosan, as these compounds are mainly generated by exhaust emissions from local sources.
Finally taking into account the recommendations established in the European guide for estimating sources of air pollution, a Factor Analysis with Varimax rotation was performed, only to determine a preliminary overview of the possible sources of the compounds studied for the Centro site [51].
Three factors were extracted that explained 84.7% of the total variance ( Table 8). The first factor, accounting for 30.7% of the variance, was mainly constituted by glutaric, malonic, and succinic acids and temperature, suggesting photochemical formation, possibly by precursors from vehicle exhausts. The second component explained 25.45% of the variance, clearly indicating local vehicle emissions. The third factor explained 28.55%, conformed mainly by K + and levoglucosan, indicating the incidence of common sources, i.e., the burning of biomass (Table 8).

Conclusions
This is the first study carried out in Cuernavaca in which the concentration of molecular markers associated with the burning of biomass was determined. It is clearly observed how, in the fire season, the concentration of such compounds is significantly increased, which contributes significantly to the deterioration of air quality in this zone and has possible implications for climate change.
The impact of biomass burning was observed by means of tracers such as levoglucosan and K + . The results showed that the high PM 2.5 concentrations and molecular tracers coincided with the fire season. The C 3 /C 4 ratio observed in this study suggested that, in the urban area of Cuernavaca, photochemical processes regulate, to a large extent, the presence of the dicarboxylic acids studied. Likewise, the K + /levoglucosan ratio reveals the relevance of open fires in the region's air quality.
The Factor Analysis gave us an overview of the possible contribution of sources of the compounds studied in the region, however, to make a more precise estimate of the possible sources, it is required to apply more advanced models, such as PMF and CMB. These software programs will be applied in the next stage of the project, in which more sampling sites and more samples will be considered.
Finally, these results constitute a call to the environmental authorities for attention to the implementation of environmental education campaigns, as well as for the generation of strategies that allow for prevention, to a large extent, of the indiscriminate burning of plant material. Funding: Authors would like to thank Consejo Nacional de Ciencia y Tecnología (CONACYT) for the studentship and Laboratorio Nacional de Estructura de Macromoléculas (LANEM CONACyT/251613).