Use of Combined Observational- and Model-Derived Photochemical Indicators to Assess the O3-NOx-VOC System Sensitivity in Urban Areas

Tropospheric levels of O3 have historically exceeded the official annual Mexican standards within the Monterrey Metropolitan Area (MMA) in NE Mexico. High-frequency and high-precision measurements of tropospheric O3, NOy, NO2, NO, CO, SO2, PM10 and PM2.5 were made at the Obispado monitoring site near the downtown MMA from September 2012 to August 2013. The seasonal cycles of O3 and NOy are driven by changes in meteorology and to a lesser extent by variations in primary emissions. The NOy levels were positively correlated with O3 precursors and inversely correlated with O3 and wind speed. Recorded data were used to assess the O3-Volatile Organic Compounds (VOC)-NOx system’s sensitivity through an observational-based approach. The photochemical indicator O3/NOy was derived from measured data during the enhanced O3 production period (12:00–18:00 Central Daylight Time (CDT), GMT-0500). The O3/NOy ratios calculated for this time period showed that the O3 production within the MMA is VOC sensitive. A box model simulation of production rates of HNO3 (PHNO3) and total peroxides (Pperox) carried out for O3 episodes in fall and spring confirmed the VOC sensitivity within the MMA environment. No significant differences were observed in O3/NOy from weekdays to weekends or for PHNO3/Pperox ratios, confirming the limiting role of VOCs in O3 production within the MMA. The ratified photochemical regime observed may allow the environmental authorities to revise and verify the current policies for air quality control within the MMA.


Introduction
Increased tropospheric levels of O 3 can be harmful for human health, vegetation and built infrastructure [1][2][3]. In the troposphere, O 3 is produced by photochemical reactions between volatile organic compounds (VOCs) and nitrogen oxides (NO x = NO + NO 2 ) in a non-linear O 3 -VOC-NO x system not fully unraveled yet [3]. Due to the non-linearity of the O 3 -VOC-NO x system, O 3 production can be VOC sensitive when controlled by the input of VOCs and increase in response to increased VOC emissions, but constant NO x levels. Conversely, O 3 production can be NO x -sensitive when NO x emissions govern the system, and O 3 mixing ratios increase in response to increased NO x emissions, but remain constant to variations of VOCs [4][5][6]. Typical VOC/NO x ratios for VOC-sensitive regimes are <4, while those for NO x -sensitive regimes are >15 [5]. However, existing studies report changes in O 3 production during the daytime and from weekdays to weekends from VOC-to NO x -sensitive regimes and vice versa within the same region as a result of changes in the emissions of precursors and meteorology [7][8][9][10][11]. Because the majority of existing policies to reduce the tropospheric levels of O 3 within urban areas focus on reducing the emissions of precursors, their success depends 1  The observational methods to assess the O 3 production sensitivity based on datasets of robust measurements for involved species in the O 3 -VOC-NO x system represent a feasible alternative to the traditional modeling approach. One advantage of this approach over pure modeling studies is that larger time frames (several months or more worth of data) can be used ( Table 1). Some of the typical photochemical indicators used in the observational approach are hydrogen peroxide (H 2 O 2 ) [5,18,20], nitric acid (HNO 3 ) [5,18,20], total odd nitrogen (NO y = NO + NO 2 + peroxyacetyl nitrate (PAN) + HNO 3 + other inorganic and organic nitrates) [12,21,22] and the O 3 /NO y ratio [23][24][25]. For example, from a numerical assessment conducted for six polluted regions in the U.S., O 3 /NO y ratios ≤ 6 and ≥8 were determined for VOC-and NOx-sensitive regimes, respectively, when mixing ratios of O 3 are >100 ppb; and O 3 /NO y ratios ≤ 11 and ≥15 in VOC-and NO x -sensitive conditions, respectively, in environments of O 3 mixing ratios < 80 ppb [25]. In Southern Taiwan, two VOC-sensitive urban areas with O 3 /NO y ratios < 6 and one NO x -sensitive area with O 3 /NO y ratios > 7 were observed during 2003-2004 [18]. O 3 /NO y average ratios of 5.1 ± 3.2 and 13.6 ± 4.7 for VOC-and partially NOx-sensitive O 3 production, respectively, at two sites in Valencia, Spain, were observed during 2010-2011 [26]. The observed variations in the O 3 /NO y ratios both for VOC-and NOx-sensitive regimes arise from different behaviors of the indicator relative to the environmental conditions (clean, moderately polluted or highly polluted environments) [25]. The photochemical indicators that have been used arise from the analysis of the main reaction pathways of the O 3 -VOC-NO x system. In a simplified manner, the initial steps of the oxidation of VOCs in the atmosphere can be represented by the following reactions [25]: (where R is a general hydrocarbons chain) (where R' is an intermediate VOC) HO • + NO 2 → HNO 3 (7) O 3 accumulates in the atmosphere as NO is transformed to NO 2 (through Reaction 2) without destroying O 3 ; i.e., as the NO + O 3 → NO 2 + O 2 reaction becomes less relevant because of the presence of VOCs that provide a source of odd hydrogen radicals that foster other reactions. Thus, the prevalence of a given photochemical regime is driven by the chemistry of hydrogen radicals. For example, it has been shown that the chain terminating steps that involve the formation of peroxides (Reactions 5 and 6) and nitric acid (Reaction 7) compete as radical sinks, and if peroxides dominate, then a NO x -sensitive condition will occur [5,25]. Similar theoretical arguments are provided to justify the use of NO y as a photochemical indicator: the split between regimes can be established from the strength of odd nitrogen sources against odd hydrogen sources. Further details can be found elsewhere [5,25].
In Mexico City, a VOC-sensitive regime was determined for most of the urban area using the O 3 /NO y , O 3 /NO z and NO y indicators derived from tropospheric measurements made at three monitoring sites within the city [12]. In addition, a numerical simulation carried out by a 3D photochemistry/transport model was used to estimate the transition values of the indicators between regimes for a two-week period in April 2004; the transition value was~8.1, and the average O 3 /NO y at the studied site was 2.6. Besides Mexico City, other large metropolitan areas in the country also experience frequent O 3 episodes, although they have received relatively little attention. For instance, the Monterrey Metropolitan Area (MMA; Figure 1), which is the third-largest metropolitan area in Mexico, has historically experienced high levels of O 3 , PM 10  At the Obispado monitoring site (OBI) located near the downtown area of Monterrey (Figure 1), the O 3 1 h NOM was exceeded annually between two and 17 times during 2000-2013, whereas the O 3 running 8 h NOM was breached between four and 38 times during the same period. Furthermore, an increase in the frequency of breaches of both the O 3 1 h average and the O 3 running 8 h average is expected due to the introduction of lower standard values of 95 and 70 ppb, respectively, applicable since October 2014. This highlights the importance of untangling the O 3 production sensitivity system to introduce effective emission controls, which can lead to an improvement in the air quality within the MMA. To date, only one study has recently assessed the O 3 production sensitivity system within the MMA [19]; a VOC-sensitive regime was observed based on numerical simulations performed with the Community Multi-scale Air Quality (CMAQ) model. However, those results come from a six-day This study presents the assessment of the O3 photochemical production regime within the MMA over a one-year period carried out by combining box-modeling and observational approaches to analyze the behavior of two photochemical indicators. The photochemical indicator O3/NOy was derived from recorded data within the MMA for tropospheric air pollutants from September 2012-August 2013, which was used to analyze the O3 production system. Ratios of the HNO3 and total peroxide production rates (PHNO3/Pperox) were computed using a box-model and were subsequently employed to evaluate the results derived from the recorded ambient data. Additionally, the existence of a weekend effect in the O3 production within the MMA was evaluated using the O3/NOy and PHNO3/Pperox ratios.

Study Site Description and Air Pollutant Monitoring
The MMA is located in northeast Mexico, some 230 km S of the United States border, and lies at an average altitude of 550 m a.s.l. (Figure 1). It is the third-most populous urban area in the country with around 5.12 million inhabitants and the second-largest industrial region [28]. The MMA also has the highest vehicle motorization index in Mexico of around 0.5 vehicles per inhabitant. Continuous measurements of typical criteria air pollutants (O3, NO, NO2, CO, SO2 and PM10) and meteorological parameters (wind speed (WS), wind direction (WD), relative humidity (RH), pressure, solar radiation (SR) and temperature) have been made since November 1992 at five monitoring sites that form part of the Integral Environmental Monitoring System (SIMA) of the Nuevo Leon Government. PM2.5 measurements began in 2003.
Additionally, NOy measurements were conducted at the OBI site from July 2012 to August 2013 using a Thermo Scientific chemiluminescence analyzer 42i-Y, in accordance with the United States Environmental Protection Agency (EPA), RFNA-1289-074. The OBI site location near the MMA downtown (25°40′33″ N, 100°20′18″ W; Figure 1) is ideal to record emissions from the industrial, domestic and mobile sources depending on air masses' trajectories. Table 2 shows the instrumentation used to measure air pollutants and meteorological parameters at OBI. Calibration This study presents the assessment of the O 3 photochemical production regime within the MMA over a one-year period carried out by combining box-modeling and observational approaches to analyze the behavior of two photochemical indicators. The photochemical indicator O 3 /NO y was derived from recorded data within the MMA for tropospheric air pollutants from September 2012 to August 2013, which was used to analyze the O 3 production system. Ratios of the HNO 3 and total peroxide production rates (P HNO3 /P perox ) were computed using a box-model and were subsequently employed to evaluate the results derived from the recorded ambient data. Additionally, the existence of a weekend effect in the O 3 production within the MMA was evaluated using the O 3 /NO y and P HNO3 /P perox ratios.

Study Site Description and Air Pollutant Monitoring
The MMA is located in northeast Mexico, some 230 km S of the United States border, and lies at an average altitude of 550 m a.s.l. (Figure 1). It is the third-most populous urban area in the country with around 5.12 million inhabitants and the second-largest industrial region [28]. The MMA also has the highest vehicle motorization index in Mexico of around 0.5 vehicles per inhabitant. Continuous measurements of typical criteria air pollutants (O 3 , NO, NO 2 , CO, SO 2 and PM 10 ) and meteorological parameters (wind speed (WS), wind direction (WD), relative humidity (RH), pressure, solar radiation (SR) and temperature) have been made since November 1992 at five monitoring sites that form part of the Integral Environmental Monitoring System (SIMA) of the Nuevo Leon Government. PM 2.5 measurements began in 2003.
Additionally, NO y measurements were conducted at the OBI site from July 2012 to August 2013 using a Thermo Scientific chemiluminescence analyzer 42i-Y, in accordance with the United States Environmental Protection Agency (EPA), RFNA-1289-074. The OBI site location near the MMA  Figure 1) is ideal to record emissions from the industrial, domestic and mobile sources depending on air masses' trajectories. Table 2 shows the instrumentation used to measure air pollutants and meteorological parameters at OBI. Calibration and maintenance procedures were carried out according to official protocols established in the Mexican standards NOM-036-SEMARNAT-1993 and NOM-156-SEMARNAT-2012.

Meteorology at the MMA
The climate at the MMA is semi-arid, with an annual average temperature of around 23 °C. Figure 3a shows that at OBI, the monthly temperature averages in summer are higher than 25 °C,

Meteorology at the MMA
The climate at the MMA is semi-arid, with an annual average temperature of around 23 • C. Figure 3a shows that at OBI, the monthly temperature averages in summer are higher than 25 • C, whereas temperatures in winter are typically below 20 • C. Similarly, the SR exhibits the highest monthly averages in summer and the lowest ones by late fall-early winter. The RH varies drastically during the year, with the lowest and highest averages typically observed in spring and fall, respectively. The rainfall within the MMA is frequent by late summer-early fall and scarce in the winter [29]. Figure 3b shows the frequency of the counts of WD occurrence at OBI by season from September 2012 to August 2013. Overall, the predominant WD is NE with frequencies of 34%-38% in spring and summer, respectively, although in winter, the greatest frequency of around 26% is observed for the E sector. Calm conditions (1 h averages) with a WS of less than 0.36 km·h −1 occurred <1% of the time. A high WS (>15 km·h −1 ) is typically observed in spring and summer for the E sector with frequencies > 2% of the time. By contrast, a WS < 3 km·h −1 shows the highest frequency in winter, mostly for the W and SW sectors. whereas temperatures in winter are typically below 20 °C. Similarly, the SR exhibits the highest monthly averages in summer and the lowest ones by late fall-early winter. The RH varies drastically during the year, with the lowest and highest averages typically observed in spring and fall, respectively. The rainfall within the MMA is frequent by late summer-early fall and scarce in the winter [29]. Figure 3b shows the frequency of the counts of WD occurrence at OBI by season from September 2012 to August 2013. Overall, the predominant WD is NE with frequencies of 34%-38% in spring and summer, respectively, although in winter, the greatest frequency of around 26% is observed for the E sector. Calm conditions (1 h averages) with a WS of less than 0.36 km·h −1 occurred <1% of the time. A high WS (>15 km·h −1 ) is typically observed in spring and summer for the E sector with frequencies > 2% of the time. By contrast, a WS < 3 km·h −1 shows the highest frequency in winter, mostly for the W and SW sectors.

Statistical Analyses
To have a better understanding of the O3/NOy photochemical indicator data obtained here, statistical tests were performed to analyze and interpret the observed pollutants' dynamics. Descriptive statistics and seasonal profiles of data recorded at OBI from September 2012 to August 2013 were calculated using the openair package [30] for R software [31]. Correlations among air pollutants and meteorological data recorded were tested using a multiple linear regression analysis. A principal component analysis (PCA) was carried out to isolate the variables that govern the O3 production within the MMA. The variables identified as drivers of the O3 production were subject to a cluster analysis (CA) to select daytime periods of enhanced photochemical activity. O3/NOy ratios calculated for the selected photochemical periods were analyzed by season, and a wind sector analysis was carried out to identify spatial variations in the photochemical processing of air masses arriving at OBI during September 2012-August 2013. Finally, the presence of a weekend effect in the diurnal production of O3 was tested using an ANOVA analysis for O3/NOy ratios during the enhanced photochemical activity period. Data correlations, PCA, CA and ANOVA were carried out using IBM SPSS Statistics software v. 19.0 (IBM, Armonk, NY, USA) for Windows.

Box Model Description and Simulations
Ratios of the HNO3 and total peroxide production rates (PHNO3/Pperox) during high photochemical activity periods can be used to assess the O3 production regime [25]. For example, typical

Statistical Analyses
To have a better understanding of the O 3 /NO y photochemical indicator data obtained here, statistical tests were performed to analyze and interpret the observed pollutants' dynamics. Descriptive statistics and seasonal profiles of data recorded at OBI from September 2012 to August 2013 were calculated using the openair package [30] for R software [31]. Correlations among air pollutants and meteorological data recorded were tested using a multiple linear regression analysis. A principal component analysis (PCA) was carried out to isolate the variables that govern the O 3 production within the MMA. The variables identified as drivers of the O 3 production were subject to a cluster analysis (CA) to select daytime periods of enhanced photochemical activity. O 3 /NO y ratios calculated for the selected photochemical periods were analyzed by season, and a wind sector analysis was carried out to identify spatial variations in the photochemical processing of air masses arriving at OBI during September 2012-August 2013. Finally, the presence of a weekend effect in the diurnal production of O 3 was tested using an ANOVA analysis for O 3 /NO y ratios during the enhanced photochemical activity period. Data correlations, PCA, CA and ANOVA were carried out using IBM SPSS Statistics software v. 19.0 (IBM, Armonk, NY, USA) for Windows.

Box Model Description and Simulations
Ratios of the HNO 3 and total peroxide production rates (P HNO3 /P perox ) during high photochemical activity periods can be used to assess the O 3 production regime [25]. For example, typical VOC-sensitive regimes exhibit P HNO3 /P perox ratios > 2. For periods of enhanced photochemical activity within the MMA from September 2012 to August 2013, P HNO3 and P perox were calculated using the California/Carnegie Institute of Technology (CIT) 3D air quality model [32][33][34] in a box-model configuration [35]. Hourly-average P HNO3 /P perox ratios were calculated from the reaction rates constants estimated by the CIT model (i.e., the SAPRC90 photochemical mechanism [36]): where P x is the rate of the production of pollutant x and k i and k j are the reaction rate constants for the corresponding production and consumption reactions, respectively. Thus, The model domain comprised a box of 16 km 2 , centered at the OBI site ( Figure 1). The model vertical structure is analogous to that used by Young et al. [35] to account for the evolution of the mixing layer, setting the top of the domain at 3100 m a.g.l. Emissions data were obtained from the National Emissions Inventory of Mexico 2005 (NEI) [37]. Emission rates for CO, NO x , VOCs, SO x and NH 3 were derived following the methodology reported by Mendoza and García [38] to obtain temporally-distributed and chemically-speciated emission rates. The chemical speciation profiles for NOx and VOC emissions were obtained from the U.S. EPA SPECIATE database for sources of emissions included in the NEI [39]. Meteorology inputs (temperature, humidity, WS and WD, SR and mixing layer height) were derived from 1-h average data recorded at the OBI site. SIMA 1-h averages of CO, NO, NO 2 , O 3 and SO 2 , together with 4-h average diurnal data for reactive hydrocarbons (RHCs), ketone, formaldehyde, acetaldehyde and isoprene data were used to constrain the model. The average RHC and individual VOC species data were obtained during sampling campaigns carried out within the MMA in the spring and fall of 2011 and 2012 [40,41]. Additionally, the CIT model was modified to include speciated NO z (NO y -NO x ) data in the initial conditions, which was calculated from NO x and NO y measurements made at the OBI site. NO z was speciated using the average contributions of the 3 main species that typically form most of the NO z produced in urban centers: 55% HNO 3 , 40% PAN and 5% nitrous acid (HONO) [42,43]. Table 3 shows the modeled periods, which were chosen because the O 3 levels breached the 110 ppb 1 h NOM applicable during 2012-2013. The modeled periods include weekends and weekdays in the fall and spring, when O 3 typically exceeds the official air quality standards.

Air Pollutants Annual Profiles
The recorded air pollutants exhibit an annual profile as a result of changes in precursor emissions and meteorology. Figure 4 shows the annual profile for O 3  with NO 2 as a result of the reduced SR and low temperatures (Figure 3a). A downward spike in the O 3 mixing ratios is observed by mid-summer 2013, which is likely caused by a high WS typical of early summer. The decrease in O 3 during summer causes another peak in the annual cycle, which is observed by early fall. However, frequent rainfall leads to lower monthly averages of O 3 during fall than those in spring. The highest mixing ratios of NO y are observed during winter as a result of the low SR and low temperature, which increase the NO 2 and NO build-up [23]. In contrast, the lowest mixing ratios of NO y are observed during summer due to enhanced dispersion, large mixing height depths and high photochemical activity of O 3 precursors. a high WS typical of early summer. The decrease in O3 during summer causes another peak in the annual cycle, which is observed by early fall. However, frequent rainfall leads to lower monthly averages of O3 during fall than those in spring. The highest mixing ratios of NOy are observed during winter as a result of the low SR and low temperature, which increase the NO2 and NO build-up [23].
In contrast, the lowest mixing ratios of NOy are observed during summer due to enhanced dispersion, large mixing height depths and high photochemical activity of O3 precursors.  Table 4 shows the results of linear correlation analyses between NOy and NO2, NO, O3, CO, SR, temperature and WS. A strong correlation between NO2 and NOy (R = 0.819) is observed in winter as a result of the low photolysis rates of NO2, which suggests that NO2 is the main component of NOy. Figure 5 shows the annual profiles of the NOy−NO difference and the NOx/NOy ratio. The NOy−NO difference exhibits the maxima and minima in winter and summer, respectively, which are in antiphase with the observed mixing ratios of O3 from September 2012-August 2013. In contrast, the NOx/NOy ratio exhibits the maxima by late fall-early winter and minima in early spring. The high values in the NOx/NOy ratio observed during winter confirm a build-up of NO2 and NO, which implies that NOz is a low fraction of the total NOy due to low photochemical processing [22]. During  Table 4 shows the results of linear correlation analyses between NO y and NO 2 , NO, O 3 , CO, SR, temperature and WS. A strong correlation between NO 2 and NO y (R = 0.819) is observed in winter as a result of the low photolysis rates of NO 2 , which suggests that NO 2 is the main component of NO y . Figure 5 shows the annual profiles of the NO y −NO difference and the NO x /NO y ratio. The NO y −NO difference exhibits the maxima and minima in winter and summer, respectively, which are in antiphase with the observed mixing ratios of O 3 from September 2012 to August 2013. In contrast, the NO x /NO y ratio exhibits the maxima by late fall-early winter and minima in early spring. The high values in the NO x /NO y ratio observed during winter confirm a build-up of NO 2 and NO, which implies that NO z is a low fraction of the total NO y due to low photochemical processing [22]. During spring, low NO x /NO y ratios indicate an enhancement of the photochemical processing of the air masses, which could confirm the high O 3 mixing ratios observed in the season (Figure 4). The presumed high contribution of NO z to NO y during spring could also explain the weak correlation between NO y and NO 2 (R 2 < 0.411) and between NO y and NO (R 2 < 0.275) ( Table 4). Finally, very weak correlations between NO y and O 3 (R 2 < 0.151) are seen during the whole year due to their antiphase annual cycle, this is underlined during spring and summer when the photo-dissociation of NO 2 to produce O 3 is enhanced [44]. spring, low NOx/NOy ratios indicate an enhancement of the photochemical processing of the air masses, which could confirm the high O3 mixing ratios observed in the season (Figure 4). The presumed high contribution of NOz to NOy during spring could also explain the weak correlation between NOy and NO2 (R 2 < 0.411) and between NOy and NO (R 2 < 0.275) ( Table 4). Finally, very weak correlations between NOy and O3 (R 2 < 0.151) are seen during the whole year due to their antiphase annual cycle, this is underlined during spring and summer when the photo-dissociation of NO2 to produce O3 is enhanced [44].   Figure 6a shows pollution roses of the O3 mixing ratios at OBI by WS. Overall, the mixing ratios of O3 > 50 ppb are frequent in air masses arriving from the NE and E sectors at a WS > 5 km·h −1 and an increase in frequency at a WS > 10 km·h −1 , likely due to the local transport of O3 and precursors  Figure 6a shows pollution roses of the O 3 mixing ratios at OBI by WS. Overall, the mixing ratios of O 3 > 50 ppb are frequent in air masses arriving from the NE and E sectors at a WS > 5 km·h −1 and an increase in frequency at a WS > 10 km·h −1 , likely due to the local transport of O 3 and precursors from the upwind dense industrial area [40,45]. Similar to the MMA, an increase in the O 3 mixing ratios caused by upwind precursor emissions was observed at the Shangdianzi site near Beijing, China [22]. In that site, it was observed that large emissions of VOCs enhance the production of O 3 linked with an increase in the NO z levels. Such an increase in NO z levels (>50 ppb) is also observed at OBI when the WS ranged from 1-5 km·h −1 for all wind sectors (Figure 6b), and for a WS > 5 km·h −1 in air masses from the N-NE-E sectors, the location of major industrial sources of NO x and VOC emissions.

Wind Sector Analysis
In contrast, mixing ratios of O 3 < 25 ppb at OBI are typical during winter and show the highest frequency at a WS < 1 km·h −1 and a reduced frequency at a WS < 5 km·h −1 for all wind sectors, except for NE and E sectors. Similar to O 3 , the NO z exhibits low mixing ratios (<25 ppb) at a WS > 10 km·h −1 ; however, at a low WS, mixing ratios of NO z > 50 ppb are common for the SW and E sectors, which is likely due to the photochemical processing of NO x emissions from mobile sources under stagnant conditions.
Atmosphere 2017, 8, 22 10 of 18 from the upwind dense industrial area [40,45]. Similar to the MMA, an increase in the O3 mixing ratios caused by upwind precursor emissions was observed at the Shangdianzi site near Beijing, China [22]. In that site, it was observed that large emissions of VOCs enhance the production of O3 linked with an increase in the NOz levels. Such an increase in NOz levels (>50 ppb) is also observed at OBI when the WS ranged from 1-5 km·h −1 for all wind sectors (Figure 6b), and for a WS > 5 km·h −1 in air masses from the N-NE-E sectors, the location of major industrial sources of NOx and VOC emissions.
In contrast, mixing ratios of O3 < 25 ppb at OBI are typical during winter and show the highest frequency at a WS < 1 km·h −1 and a reduced frequency at a WS < 5 km·h −1 for all wind sectors, except for NE and E sectors. Similar to O3, the NOz exhibits low mixing ratios (<25 ppb) at a WS > 10 km·h −1 ; however, at a low WS, mixing ratios of NOz > 50 ppb are common for the SW and E sectors, which is likely due to the photochemical processing of NOx emissions from mobile sources under stagnant conditions.

The Enhanced Photochemical Period
Photochemical indicators around the period of maximum photochemical activity for the chemical species involved in the O3 production system may reflect daytime variations in photochemistry within the planetary boundary layer, and therefore, such indicators can be used to assess the photochemical regime of O3 production [5,26]. For example, O3/NOy ratios were estimated from data recorded in the period of 13:00-17:00 Central Daylight Time (CDT, GMT-0500) during April 2004 at a downwind receptor site of photo-chemically-aged air masses within Mexico City [12]. Likewise, O3/NOy ratios from measurements made during 13:00-16:00 CET at two sampling sites in Valencia, Spain, during August 2010-May 2011 and May-October 2011 were calculated [26]. In the current study, the variables that govern the O3 production within the MMA were isolated using a PCA for NOy, NO2, NO, O3, CO, SO2, PM10, PM2.5 and temperature, WS, WD and SR data recorded from September 2012 to August 2013. Table 5 shows that three components designated as PC1-3 are significant, which accounted for 67.2% of the total variability. The PC1 revealed a positive correlation among the precursors of O3; NOy, NO, NO2 and CO. The PC2 correlates positively with O3 and SR, which is explained by the

The Enhanced Photochemical Period
Photochemical indicators around the period of maximum photochemical activity for the chemical species involved in the O 3 production system may reflect daytime variations in photochemistry within the planetary boundary layer, and therefore, such indicators can be used to assess the photochemical regime of O 3 production [5,26]. For example, O 3 /NO y ratios were estimated from data recorded in the period of 13:00-17:00 Central Daylight Time (CDT, GMT-0500) during April 2004 at a downwind receptor site of photo-chemically-aged air masses within Mexico City [12]. Likewise, O 3 /NO y ratios from measurements made during 13:00-16:00 CET at two sampling sites in Valencia, Spain, during August 2010-May 2011 and May-October 2011 were calculated [26]. In the current study, the variables that govern the O 3 production within the MMA were isolated using a PCA for NO y , NO 2 , NO, O 3 , CO, SO 2 , PM 10 , PM 2.5 and temperature, WS, WD and SR data recorded from September 2012 to August 2013. Table 5 shows that three components designated as PC1-3 are significant, which accounted for 67.2% of the total variability. The PC1 revealed a positive correlation among the precursors of O 3 ; NO y , NO, NO 2 and CO. The PC2 correlates positively with O 3 and SR, which is explained by the photolysis of NO 2 during the daytime. The PC3 correlates positively with WD and temperature, which comprise the effect of the air mass origin and planetary boundary layer height that influences the dispersion of O 3 . Dendrograms for the O 3 and SR data recorded within the MMA (PC2) were constructed to identify the hours of enhanced photochemistry, the period of maximum O 3 production. Figure 7 shows the annual average period of enhanced O 3 production from 12:00 to 18:00 CDT and O 3 depletion from 19:00 to 11:00, respectively. photolysis of NO2 during the daytime. The PC3 correlates positively with WD and temperature, which comprise the effect of the air mass origin and planetary boundary layer height that influences the dispersion of O3. Dendrograms for the O3 and SR data recorded within the MMA (PC2) were constructed to identify the hours of enhanced photochemistry, the period of maximum O3 production. Figure 7 shows the annual average period of enhanced O3 production from 12:00-18:00 CDT and O3 depletion from 19:00-11:00, respectively.

Use of the O3/NOy Photochemical Indicator
The photochemical indicator O3/NOy was derived from measurements made during the period of 12:00-18:00 CDT at OBI during September 2012-August 2013. Figure 8 shows a box plot by season for O3/NOy ratios at OBI. Overall, the O3/NOy ratio ranged from 0.1 in fall 2012 to 4.8 in summer 2013, while medians and averages in O3/NOy ranged from 0.8 and from 0.9 in winter 2012 to 2.2 and to 2.3 in summer, respectively. The low O3/NOy ratios calculated at OBI during winter result from low O3 levels and high NOy levels, whereas the high O3/NOy ratios during summer derive from moderate O3 levels, but low NOy levels. The O3/NOy ratios observed in all seasons suggest that the O3 production within the MMA is VOC sensitive throughout the entire year. This is in good agreement with prior results of

Use of the O 3 /NO y Photochemical Indicator
The photochemical indicator O 3 /NO y was derived from measurements made during the period of 12:00-18:00 CDT at OBI during September 2012-August 2013. Figure 8 shows a box plot by season for O 3 /NO y ratios at OBI. Overall, the O 3 /NO y ratio ranged from 0.1 in fall 2012 to 4.8 in summer 2013, while medians and averages in O 3 /NO y ranged from 0.8 and from 0.9 in winter 2012 to 2.2 and to 2.3 in summer, respectively. The low O 3 /NO y ratios calculated at OBI during winter result from low O 3 levels and high NO y levels, whereas the high O 3 /NO y ratios during summer derive from moderate O 3 levels, but low NO y levels. The O 3 /NO y ratios observed in all seasons suggest that the O 3 production within the MMA is VOC sensitive throughout the entire year. This is in good agreement with prior results of O 3 production being VOC sensitive within the MMA based on numerically-modeled O 3 /NO y ratios that ranged between 2.9 and when the highest 3.5 for the OBI site [19] and that are within the range of those calculated in the current study. The differences observed between the current and the referred prior results may arise from the fact that their modeled ratios were provided exclusively for 13:00 CDT and limited to a pollution episode in summer 2005.
Atmosphere 2017, 8, 22 12 of 18 O3 production being VOC sensitive within the MMA based on numerically-modeled O3/NOy ratios that ranged between 2.9 and when the highest 3.5 for the OBI site [19] and that are within the range of those calculated in the current study. The differences observed between the current and the referred prior results may arise from the fact that their modeled ratios were provided exclusively for 13:00 CDT and limited to a pollution episode in summer 2005. The O3/NOy ratios calculated from 12:00-18:00 CDT at OBI were used to construct pollution roses by WS, which are shown in Figure 9. Overall, O3/NOy ratios < 2 are predominant at a WS < 5 km·h −1 for all wind sectors and increased proportionally to WS, at a WS > 5 km·h −1 , although, this is only seen for the NE, E and SE sectors. The highest O3/NOy ratios observed (>4) were recorded in air masses arriving from the easterly sectors, NE-E-SE at a WS > 5 km·h −1 , and exhibit the highest frequency in line with the highest WS observed. Low O3/NOy ratios are typical in winter as result of low temperatures and low WS occurrence. A low WS may influence the formation and local transport of O3 by limiting the horizontal and vertical mixing and the reactions of O3 precursor emissions, which can also occur during other seasons. Moreover, at a low WS, the observed high values of NOx/NOy in Figure 5 suggest a low contribution of NOz to NOy, which is typical of low photochemical processing commonly seen in VOC-sensitive regimes [46]. By contrast, a high WS may enhance the photochemical processing of O3 precursors and the transport of air masses travelling over rural areas located east of the MMA that typically have NOx-sensitive regimes.  The O 3 /NO y ratios calculated from 12:00 to 18:00 CDT at OBI were used to construct pollution roses by WS, which are shown in Figure 9. Overall, O 3 /NO y ratios < 2 are predominant at a WS < 5 km·h −1 for all wind sectors and increased proportionally to WS, at a WS > 5 km·h −1 , although, this is only seen for the NE, E and SE sectors. The highest O 3 /NO y ratios observed (>4) were recorded in air masses arriving from the easterly sectors, NE-E-SE at a WS > 5 km·h −1 , and exhibit the highest frequency in line with the highest WS observed. Low O 3 /NO y ratios are typical in winter as result of low temperatures and low WS occurrence. A low WS may influence the formation and local transport of O 3 by limiting the horizontal and vertical mixing and the reactions of O 3 precursor emissions, which can also occur during other seasons. Moreover, at a low WS, the observed high values of NO x /NO y in Figure 5 suggest a low contribution of NO z to NO y , which is typical of low photochemical processing commonly seen in VOC-sensitive regimes [46]. By contrast, a high WS may enhance the photochemical processing of O 3 precursors and the transport of air masses travelling over rural areas located east of the MMA that typically have NO x -sensitive regimes.
Atmosphere 2017, 8, 22 12 of 18 O3 production being VOC sensitive within the MMA based on numerically-modeled O3/NOy ratios that ranged between 2.9 and when the highest 3.5 for the OBI site [19] and that are within the range of those calculated in the current study. The differences observed between the current and the referred prior results may arise from the fact that their modeled ratios were provided exclusively for 13:00 CDT and limited to a pollution episode in summer 2005. The O3/NOy ratios calculated from 12:00-18:00 CDT at OBI were used to construct pollution roses by WS, which are shown in Figure 9. Overall, O3/NOy ratios < 2 are predominant at a WS < 5 km·h −1 for all wind sectors and increased proportionally to WS, at a WS > 5 km·h −1 , although, this is only seen for the NE, E and SE sectors. The highest O3/NOy ratios observed (>4) were recorded in air masses arriving from the easterly sectors, NE-E-SE at a WS > 5 km·h −1 , and exhibit the highest frequency in line with the highest WS observed. Low O3/NOy ratios are typical in winter as result of low temperatures and low WS occurrence. A low WS may influence the formation and local transport of O3 by limiting the horizontal and vertical mixing and the reactions of O3 precursor emissions, which can also occur during other seasons. Moreover, at a low WS, the observed high values of NOx/NOy in Figure 5 suggest a low contribution of NOz to NOy, which is typical of low photochemical processing commonly seen in VOC-sensitive regimes [46]. By contrast, a high WS may enhance the photochemical processing of O3 precursors and the transport of air masses travelling over rural areas located east of the MMA that typically have NOx-sensitive regimes.  No weekend effect (significant differences, p > 0.05) was observed in the O 3 mixing ratios and O 3 /NO y ratios between weekdays and weekends of all seasons (Table 6; Figure 10), despite the lower average O 3 /NO y ratios during weekdays. This lack of a weekend effect in the O 3 mixing ratios and the average O 3 /NO y ratios arise from the limiting role of VOCs in the weekday O 3 production, while reduced vehicular NO x emissions during weekends increase the VOC/NO y emission during weekends [47,48]. This decrease has counteracting effects on the O 3 production leading to similar O 3 mixing ratios (±5%) during weekdays and weekends, which was also reported for Mexico City between 1986 and 2007 [49]. By contrast, a weekend effect in O 3 levels was observed between 2007 and 2009 within the urban areas of Oporto and Lisbon in Portugal and London, which was ascribed to changes in meteorology [9]. Within the MMA, the lowest difference between weekdays and weekends in the O 3 /NO y ratios is seen in winter, when the lowest O 3 /NO y is also observed and contrasts with the largest difference seen in summer when the highest O 3 /NO y was calculated. The higher O 3 /NO y ratios observed during summer are likely due to a combination of low NO x emissions and meteorological conditions that foster the fast dispersion of air pollutants, limiting the presence of photochemically-processed air masses in the MMA [47].
Atmosphere 2017, 8, 22 13 of 18 No weekend effect (significant differences, p > 0.05) was observed in the O3 mixing ratios and O3/NOy ratios between weekdays and weekends of all seasons (Table 6; Figure 10), despite the lower average O3/NOy ratios during weekdays. This lack of a weekend effect in the O3 mixing ratios and the average O3/NOy ratios arise from the limiting role of VOCs in the weekday O3 production, while reduced vehicular NOx emissions during weekends increase the VOC/NOy emission during weekends [47,48]. This decrease has counteracting effects on the O3 production leading to similar O3 mixing ratios (±5%) during weekdays and weekends, which was also reported for Mexico City between 1986 and 2007 [49]. By contrast, a weekend effect in O3 levels was observed between 2007 and 2009 within the urban areas of Oporto and Lisbon in Portugal and London, which was ascribed to changes in meteorology [9]. Within the MMA, the lowest difference between weekdays and weekends in the O3/NOy ratios is seen in winter, when the lowest O3/NOy is also observed and contrasts with the largest difference seen in summer when the highest O3/NOy was calculated. The higher O3/NOy ratios observed during summer are likely due to a combination of low NOx emissions and meteorological conditions that foster the fast dispersion of air pollutants, limiting the presence of photochemically-processed air masses in the MMA [47].

Box Modeling
The Pperox was calculated by summing the production rates of H2O2 (PH2O2), hydroperoxides and other peroxides, although PH2O2 represents by far the largest contribution to Pperox. The PHNO3 was estimated from the reaction rate of the HO • + NO2 → HNO3 reaction. Figure 11a shows the distribution of the hourly average PHNO3/Pperox ratios calculated between 12:00 and 18:00 CDT for each

Box Modeling
The P perox was calculated by summing the production rates of H 2 O 2 (P H2O2 ), hydroperoxides and other peroxides, although P H2O2 represents by far the largest contribution to P perox . The P HNO3 was estimated from the reaction rate of the HO • + NO 2 → HNO 3 reaction. Figure 11a shows the distribution of the hourly average P HNO3 /P perox ratios calculated between 12:00 and 18:00 CDT for each modeled period. The modeled P HNO3 /P perox ratios are consistently >2 for all selected periods, which correspond to a VOC-sensitive regime and are in good agreement with the results derived from the observational approach described here. Negative values of P HNO3 /P perox ratios account for 5.6% of the total data and are observed mostly between 12:00 and 13:00 CDT due to peroxide consumption at the beginning of the period of enhanced O 3 production. Figure 11b shows that the O 3 /NO y ratios calculated for the whole modelling period were consistently < 6, which confirms the VOC-sensitive regime in O 3 production within the MMA suggested by the P HNO3 /P perox results.
Atmosphere 2017, 8, 22 14 of 18 modeled period. The modeled PHNO3/Pperox ratios are consistently >2 for all selected periods, which correspond to a VOC-sensitive regime and are in good agreement with the results derived from the observational approach described here. Negative values of PHNO3/Pperox ratios account for 5.6% of the total data and are observed mostly between 12:00-13:00 CDT due to peroxide consumption at the beginning of the period of enhanced O3 production. Figure 11b shows that the O3/NOy ratios calculated for the whole modelling period were consistently < 6, which confirms the VOC-sensitive regime in O3 production within the MMA suggested by the PHNO3/Pperox results. Additionally, to confirm the lack of a weekend effect in O3 production within the MMA determined from the observational approach, an ANOVA analysis was performed to compare the modeled PHNO3/Pperox and O3/NOy ratios for weekdays and weekends of the modeled periods. Table 7 shows that no significant differences (p > 0.05) are observed between the average PHNO3/Pperox ratios during weekdays and weekends and between the average O3/NOy ratios during weekdays and weekends as shown in Table 6, which confirms the limiting role of VOC in production within the MMA.  Additionally, to confirm the lack of a weekend effect in O 3 production within the MMA determined from the observational approach, an ANOVA analysis was performed to compare the modeled P HNO3 /P perox and O 3 /NO y ratios for weekdays and weekends of the modeled periods. Table 7 shows that no significant differences (p > 0.05) are observed between the average P HNO3 /P perox ratios during weekdays and weekends and between the average O 3 /NO y ratios during weekdays and weekends as shown in Table 6, which confirms the limiting role of VOC in production within the MMA. The O 3 sensitivity results presented here are relevant in the context of new energy-oriented projects that are under development in the northeast of Mexico, some of them located east of the MMA. Such projects include new natural gas combined-cycle power plants that are projected to start operations by late 2016, which could have an accumulated installed capacity of up to +1.3 GW. Currently, electric utilities already installed east of the MMA add up to a total capacity of +2.1 GW, all of those being natural gas combined-cycle plants. The new facilities are projected to be supplied with natural gas imported from Texas and with shale gas expected to be exploited from the Cuenca de Burgos Basin that is also located east of the MMA. The Burgos Basin represents around two-thirds of the estimated 550 trillion cubic feet of shale gas recoverable in Mexico, which is the sixth largest reservoir in the world [50]. The introduction of shale gas extraction and energy production will likely increase the regional emissions of VOCs and NO x , impacting the photochemistry of the MMA airshed during events of enhanced regional transport [51]. Increasing NO x levels upwind of the MMA could foster higher O 3 levels.
Finally, from the perspective of control strategies that could be put into place to help alleviate the air pollution problem that the MMA faces, it is relevant to match the results obtained here with the local emissions inventory. According to the latest comprehensive official inventory published for the MMA [37], 47% of the VOC emissions come from mobile sources and 43% from area sources; only 8% come from point sources. In contrast, the contribution of NO x emissions is led by mobile (48%) and point (33%) sources. From a mass-basis perspective, one could argue that control strategies should target VOC emissions from mobile sources, in particular light-duty vehicles, which account for more than 70% of these emissions, and area sources. For the latter, the main contributions are from domestic use of solvents (34%), surface cleaning (13%), liquefied petroleum gas (LPG) leaks (16%), building painting activities (9%), industrial painting processes (9%) and fugitive emissions from gasoline distribution and handling in service stations (9%). However, it has to be recognize that the composition of these VOC mixtures changes from source to source, making their ozone-forming potential different. Further studies are needed to compare the reactivity of these mixtures to establish the real benefits of reducing the emissions of one source or another. If NO x emission control strategies are explored, these should be accompanied with VOC control strategies to ensure O 3 reductions [19].

Conclusions
Continuous measurements of tropospheric O 3 and NO y were made at the OBI site near the downtown MMA and used to assess the sensitivity of the O 3 production system from September 2012 to August 2013. Within the MMA, O 3 exhibits maxima in spring in response to the enhanced photolysis of NO 2 , whereas the minima are observed in winter due to the reduced SR. The highest mixing ratios of O 3 were observed in easterly air masses at a WS of 5-10 km·h -1 , and the lowest ones were recorded in calm winds throughout the entire year. The NO y peaks during winter and decreases during summer, which suggests that during summer, the photochemical production is oriented to O 3 rather than to NO y . During winter, the recorded data revealed that NO and NO 2 are the major components of NO y .
The O 3 production is enhanced between 12:00 and 18:00 CDT in line with the period of maximum SR. O 3 /NO y ratios <6 were observed during the year studied, suggesting that O 3 production within the MMA is VOC sensitive. Modeled P HNO3 /P perox ratios > 2 for periods of O 3 episodes in fall and spring confirm the VOC-sensitive environment within the MMA derived from the observational analysis performed. The non-significant differences observed in O 3 /NO y and in P HNO3 /P perox between weekdays and weekends suggest the lack of a weekend effect in O 3 production. The lack of an O 3 weekend effect within the MMA confirms the limiting role of VOCs in O 3 production during weekdays. This study demonstrates the usefulness of high-precision measurements of O 3 and NO y to assess the O 3 -VOC-NO x system's sensitivity and to independently test the accuracy of box chemical models. The results presented here allow the wholly independent validation of current air quality policies directed to reduce tropospheric O 3 levels and, if required, the design and implementation of new ones.