Trends of Ground-Level Ozone in New York City Area during 2007–2017

: The spatiotemporal patterns of ground level ozone (O 3 ) concentrations in the New York City (NYC) metropolitan region for the 2007–2017 period were examined conjointly with local emissions of O 3 precursors and the frequency of wildﬁres. Daily 8-h and 1-h O 3 and nitric oxide (NO) concentrations were retrieved from the US Environmental Protection Agency (EPA) Air Data. Annual emission inventories for 2008 and 2017 were acquired from EPA National Emissions Inventory (NEI). The number and area burnt by natural and human-ignited wildﬁres were acquired from the National Interagency Fire Center (NIFC). The highest daily 8-h max O 3 concentrations varied from 90 to 111 parts per billion volume (ppbv) with the highest concentrations measured perimetrically to NYC urban agglomeration. The monthly 8-h max O 3 levels have been declining for most of the peri-urban sites but increasing (from +0.18 to +1.39 ppbv/year) for sites within the urban agglomeration. Slightly higher O 3 concentrations were measured during weekend than those measured during the weekdays in urban sites probably due to reduced O 3 titration by NO. Signiﬁcant reductions of locally emitted anthropogenic nitrogen oxides (NO x ) and volatile organic compounds (VOCs) may have triggered the transition from VOC-limited to NO X -limited conditions, with downwind VOCs sources being critically important. Strong correlations between the monthly 8-h max O 3 concentrations and wildﬁres in Eastern US were computed. More and destructive wildﬁres in the region were ignited by lightning for years with moderate and strong La Niña conditions. These ﬁndings indicate that climate change may counterbalance current and future gains on O 3 precursor’s reductions by amending the VOCs-to-NO x balance.


Introduction
Ground-level ozone (O 3 ) negatively impacts human health across life stages, natural ecosystems, and climate [1,2]. Ozone is a strong oxidative agent that reacts with proteins and lipids in the airways lining fluid of the lung and compromised lung function [3][4][5]. Early-life exposure to O 3 affects the growth and function of developing lungs and may promote the asthma phenotype in the first year of life [6][7][8]. Children exposed to O 3 are also more likely to have airway hyper responsiveness [9]. O 3 exposures have been consistently shown to increase asthma medication use, mortality, emergency department (ED) visits, and hospitalizations by exacerbating asthma and chronic obstructive pulmonary disease (COPD) [1,10]. O 3 , a secondary pollutant, is produced from the daytime oxidation of irradiated mixtures of volatile organic compounds (VOCs) and nitrogen oxides (NO x ) (NO + NO 2 = NO x ) (precursors, thereafter). O 3 levels increase as VOCs levels increase, while increasing NO x levels may either generate or titrate O 3 depending on instantaneous VOC/NO x ratio (in parts per million carbon (ppmC)/parts per million (ppm)) [11]. VOCs and NO x are emitted from fossil and contemporary fuel combustion in anthropogenic activities and wildfires.  Figure 1 and three of them in Suffolk County, Long Island (#10, #11 (500,000-700,000 people) and #12 (57,000 people) in Figure 1); • Six sites were operated by the New Jersey Department of Environmental Protection. Three of the sites (#2, #5 and #15) were in populated urban settings (from 600,000 to 2,500,000 people within 8-km radius), while the remaining three were further away from New York City (#13, #14 and #16, less than 275,000 people within 8-km); and • Three sites along the US Interstate-95 highway to Bridgeport, CT, operated by the Connecticut Department of Environmental Quality (#7, #8, and #9 in Figure 1, with 180,000 to 280,000 people living within an 8-km radius).
Ambient O 3 concentrations were photometrically measured with approved federal equivalent methods.    [20]) and primary road network in the study area. The numbers in the circle refer to sites ID# in Table 1.

Emissions Inventories and Wildfires
The 2008 and 2017 NOx and VOCs emissions for New York, New Jersey, and Connecticut were obtained from the USEPA National Emissions Inventory (NEI). It includes emissions from point, area, mobile (on-and off-road), and event-specific sources based on source activity data provided by the state, local, and tribal air agencies through the Emissions Inventory System (EIS). Emissions have been reported by EIS sectors since 2008 as described in the Source Classification Codes (SCCs). SCCs are source-specific processes  [20]) and primary road network in the study area. The numbers in the circle refer to sites ID# in Table 1.

Emissions Inventories and Wildfires
The 2008 and 2017 NO x and VOCs emissions for New York, New Jersey, and Connecticut were obtained from the USEPA National Emissions Inventory (NEI). It includes emissions from point, area, mobile (on-and off-road), and event-specific sources based on source activity data provided by the state, local, and tribal air agencies through the Emissions Inventory System (EIS). Emissions have been reported by EIS sectors since 2008 as described in the Source Classification Codes (SCCs). SCCs are source-specific processes or functions that emit air pollutants. For this study, 2008 and 2017 NEI data by EIS sectors were grouped into fourteen source types as follows: agriculture/livestock waste; biogenics (including vegetation and soil); bulk gasoline terminals; commercial cooking; fires (including agricultural, prescribed, and wildfires); commercial, electrical generation, industrial and residential fuel combustion; gas stations; industrial and non-industrial processes; mobile sources, solvent fugitive emissions; and waste disposal. The number of human-and lightning-ignited wildfires and area burnt for the 2007-2017 period by year for each of the eleven Geographic Area Coordinating Group (GACG) were obtained from the National Interagency Fire Center (NIFC) [23]. The study area is part of the Eastern Area Coordination Center (EACC) that includes a total of twenty (20)

Data Analysis
Ambient daily 8-h max O 3 concentrations were tested for normality using the Shapiro-Wilk test. The significance of difference among sites was assessed with the non-parametric Kruskal-Wallis at α = 0.05. The daily 8-h max O 3 absolute (∆C) and the percent relative (%∆C/Ref) concentration differences and the coefficient of divergence (COD) were computed [24]. The CCNY (located in City College of New York) (Site #1 in Figure 1) was the reference site because of its central location to the study area. COD values vary from 0 to 1, with high COD values being indicative of spatial gradient. The paired ∆C between two sites were used to determine whether concentrations change simultaneously among the sites over time. The %∆C/C Ref of 24-h paired concentration was computed to assess systematic differences between the sites and site-to-site variation, respectively [24]. COD was used to assess the spatial uniformity of measurements with respect to the concentration levels [25].
The monthly 8-h max O 3 concentration was computed for months with more than 75% of daily 8-h max O 3 measurements. It has been previously used to examine the effect of wildfires on O 3 , compliance with NAAQS, and to smooth the effects of local meteorology and short-term changes in local emissions [24,25]. The annual trend was computed using the de-seasonalized monthly 8-h max O 3 concentrations by applying the non-parametric sequential Mann-Kendall test at a confidence level of 95% [26,27].
Hourly NO and O 3 concentrations were used to compute the morning NO-O 3 crossover time (tNOxO3) (in h) and the O 3 accumulation time (tO3_acc) (in h) as follows: (i) the tNO X O3 is the time of the day where NO and O 3 profiles intersected after the early morning NO peak; and (ii) the tO3_acc is the time of the day with the highest O 3 concentration [11] The ozone accumulation rate (in ppbv O 3 /h) was computed as follows: The two-tailed Spearman correlation coefficient was computed to assess the relationship between wildfires and annual 8-h max O 3 concentration. Analyses were done using SPSS (Version 26) (IBM Analytics, Armonk, NY, USA) and Origin Pro (version 9.1) (Origin Lab, Northampton, MA, USA).     Table 2 shows the 2017 8-h max O 3 concentration (in ppbv), the site-specific COD value, median (and standard deviation (σ)) of absolute (∆C) and relative (%∆C/Ref) concentration differences (relative to CCNY air quality monitoring site) and the annual trends of monthly 8-h max O 3 levels for each site (ppbv/year). The 8-h maximum O 3 concentrations for 2017 varied from 60 ppbv for the coastal site upwind of New York City in Monmouth University (#16) to 81 ppbv for the two coastal sites downwind of New York City in Connecticut (Sherwood Island (#8) and Stratford (#9)). The sites were grouped in two clusters based on spatiotemporal similarities to the reference site (#1 at CCNY in Manhattan). The first cluster was composed of sites within the NYC urban area (#3-5), with low COD values (from 0.10 to 0.14), ∆C (2-4 ppbv) and %∆C/C ref (from 6 to 12%) indicating the lack of a spatial gradient within the densely populated area. For most of these sites (except for Bayonne (#5) that is located on the south of the densely NYC populated area), the 8-h max O 3 concentrations were increasing from 0.18 to 1.39 ppbv/year. For sites located perimetrically to NYC, slightly higher COD (from 0.13 to 0.22), ∆C (from 5 to 9 ppbv) and %∆C/C ref (from 16 to 26%) suggested a weak spatial trend particularly across the west-east axis relative to NYC. The 8-h max O 3 concentrations declined from −0.25 to −1.82 ppbv/year except for two of the nearest to NYC sites in Connecticut from +0.03 (#7) and +0.43 (#8) ppbv/year.  Figure 3 shows the mean (±3 × standard error) ozone weekend-to-weekday effect (OWE) ratio of average maximum O 3 concentrations for each site. OWE values higher than one indicate that weekend O 3 concentrations were higher than those measured during the weekdays. For all NYC urban sites, the OWE ratio was higher than 1 (from 1.04 to 1.07), indicating that weekend O 3 levels were higher than those measured during weekend. For two of the most populated peri-urban sites (White Plains (#6) and Babylon (#10)), the OWE ratio was also higher than 1 (1.02 and 1.03, respectively), and less than one for the remaining peri-urban sites.  Figure 4 presents the day-of-week variation of the morning NOxO3 crossover time (tNOxO3), the O3 accumulation time (tO3_acc) and the O3 accumulation rate for (a) Queens College (NYC urban site (#4)) and (b) Chester (peri-urban site (#14)). For Queens College site, the tNOxO3 crossover was observed at between 7:00 and 9:00 am and O3 accumulated for six hours during weekdays and Saturday (Figure 4a). During Sunday, O3 formation was not hampered because of low NO levels. The weekend NO concentrations (24-h mean: 1.4 −1.8 ppbv; 1-h max: 3.7-6.9 ppbv) was up to three times lower than that measured during weekdays (24-h mean: 2.7-3.4 ppbv; 1-h max: 7.7-14.0 ppbv). The lowest accumulation rate observed on Sunday (1.24 ppbv/h) was counterbalanced by the longer accumulation period, resulting in elevated O3 concentrations. Because of the lack of titration in the early morning, the O3 overnight carryover on weekend (1-h: 18.6-23.4 ppbv) was higher than that during weekdays (1-h: 13-16.7 ppbv), adding to higher O3 concentrations in weekend.      Figure 4 presents the day-of-week variation of the morning NOxO3 crossover time (tNOxO3), the O3 accumulation time (tO3_acc) and the O3 accumulation rate for (a) Queens College (NYC urban site (#4)) and (b) Chester (peri-urban site (#14)). For Queens College site, the tNOxO3 crossover was observed at between 7:00 and 9:00 am and O3 accumulated for six hours during weekdays and Saturday (Figure 4a). During Sunday, O3 formation was not hampered because of low NO levels. The weekend NO concentrations (24-h mean: 1.4 −1.8 ppbv; 1-h max: 3.7-6.9 ppbv) was up to three times lower than that measured during weekdays (24-h mean: 2.7-3.4 ppbv; 1-h max: 7.7-14.0 ppbv). The lowest accumulation rate observed on Sunday (1.24 ppbv/h) was counterbalanced by the longer accumulation period, resulting in elevated O3 concentrations. Because of the lack of titration in the early morning, the O3 overnight carryover on weekend (1-h: 18.6-23.4 ppbv) was higher than that during weekdays (1-h: 13-16.7 ppbv), adding to higher O3 concentrations in weekend.   For Chester (Figure 4b), NO levels were minimal during weekends (1-h max levels < 0.2 ppbv) and less than 1 ppbv during weekdays. As such, there was no O 3 titration, with prolonged periods of ozone accumulation albeit at low accumulation rates (0.81-1.08 ppbv/h) for both weekends and weekdays. Moreover, the O 3 overnight carryover (28.5-31.3 ppbv) did not vary among different days of the week. For Chester (Figure 4b), NO levels were minimal during weekends (1-h max levels < 0.2 ppbv) and less than 1 ppbv during weekdays. As such, there was no O3 titration, with prolonged periods of ozone accumulation albeit at low accumulation rates (0.81-1.08 ppbv/h) for both weekends and weekdays. Moreover, the O3 overnight carryover (28.5-31.3 ppbv) did not vary among different days of the week.  Table 3 shows the Spearman correlation coefficient of O3 concentrations, the number of and area burnt by fires within the areas managed by the ten GACC coordinating centers. Moderate to strong correlations were computed for the number (0.91, p < 0.001) and area (0.60, p = 0.01) burnt by lightning-ignited fires in the Eastern Area followed by fires in the Southern Area coordination center (number: 0.80, p = 0.003) encompassing all states east of the Mississippi River and adjacent westerly states. Weaker correlations were computed for human-ignited wildfires in the same regions. It is noteworthy that prescribed burns in winter and spring for ecological management and to manage biomass fuel on the Table 3 shows the Spearman correlation coefficient of O 3 concentrations, the number of and area burnt by fires within the areas managed by the ten GACC coordinating centers. Moderate to strong correlations were computed for the number (0.91, p < 0.001) and area (0.60, p = 0.01) burnt by lightning-ignited fires in the Eastern Area followed by fires in the Southern Area coordination center (number: 0.80, p = 0.003) encompassing all states east of the Mississippi River and adjacent westerly states. Weaker correlations were computed for human-ignited wildfires in the same regions. It is noteworthy that prescribed burns in winter and spring for ecological management and to manage biomass fuel on the forest floor account for most of the human-induced fires in the Southern area. The Spearman correlation coefficient declined for wildfires further away from the study area Table 3. Spearman correlation coefficients of monthly 8-h max O 3 concentrations and lightning, the number and area burnt by human-induced wildfires in regional GACG coordinating centers.

Number
Area ; Great Basin (Utah, Nevada, Idaho-south of the Salmon River, the western Wyoming mountains, and the Arizona Strip); Northwest (Washington and Oregon). ** Correlation is significant at the 0.01 level (two-tailed) *** Correlation is significant at the 0.001 level (two-tailed).
The annual variation of 8-h max O 3 concentration, number and area burnt by lightningignited wildfires in the Eastern and Southern GACG areas are presented in Figure 6a,c,d.

Discussion
In this study, we observed heterogeneity in the trends of monthly 8-h max O3 concentrations in urban and peri-urban sites in the NYC metropolitan area during 2007-2017. The declining trends in peri-urban sites are consistent with national O3 trends for less urbanized and rural areas [12]. These sites are usually located downwind of urban agglomerations where the highest 8-h max O3 concentrations were historically recorded. Conversely, an increasing trend of O3 was observed for the sites located within the urban agglomeration, areas that historically experienced high traffic-related NOx emissions and low O3 concentrations. This is also in agreement with increasing trends at urbanized locations across the US [12]. Similar trends were observed across the world with increasing concentrations in urban areas (0.31 ppbv/year) and declining O3 levels in rural areas (−0.23 ppbv/year) over the past three decades [28,29]. The opposite trends in O3 concentrations in peri-urban and urban sites can be tentatively explained by declining anthropogenic VOCs and NOx emissions over the past decades, and the non-linear sensitivity of O3 formation to VOCs-to-NOx instantaneous ratio [30]. In urban areas, O3 levels are conditioned by NO (from vehicular emission) titration in the early morning. Because of the significant declines in NOx emissions, titration of O3 by NO was reduced leading to an increase of nighttime carryover O3 [30]. Ninneman and Jaffe [21] computed that the summertime ozone production efficiency in New York State rural sites increased in response to NOx reductions in NOx-limited conditions. The VOCs-to-NOx ratio between 2008 and 2017 (based on EPA NEI) may have increased from 26.7% to 83.5% depending on VOCs composition that can transition from VOC-limited conditions to NOx-limited for O3 formation.

Discussion
In this study, we observed heterogeneity in the trends of monthly 8-h max O 3 concentrations in urban and peri-urban sites in the NYC metropolitan area during 2007-2017. The declining trends in peri-urban sites are consistent with national O 3 trends for less urbanized and rural areas [12]. These sites are usually located downwind of urban agglomerations where the highest 8-h max O 3 concentrations were historically recorded. Conversely, an increasing trend of O 3 was observed for the sites located within the urban agglomeration, areas that historically experienced high traffic-related NO x emissions and low O 3 concentrations. This is also in agreement with increasing trends at urbanized locations across the US [12]. Similar trends were observed across the world with increasing concentrations in urban areas (0.31 ppbv/year) and declining O 3 levels in rural areas (−0.23 ppbv/year) over the past three decades [28,29]. The opposite trends in O 3 concentrations in peri-urban and urban sites can be tentatively explained by declining anthropogenic VOCs and NO x emissions over the past decades, and the non-linear sensitivity of O 3 formation to VOCsto-NO x instantaneous ratio [30]. In urban areas, O 3 levels are conditioned by NO (from vehicular emission) titration in the early morning. Because of the significant declines in NO x emissions, titration of O 3 by NO was reduced leading to an increase of nighttime carryover O 3 [30]. Ninneman and Jaffe [21] computed that the summertime ozone production efficiency in New York State rural sites increased in response to NO x reductions in NO x -limited conditions. The VOCs-to-NO x ratio between 2008 and 2017 (based on EPA NEI) may have increased from 26.7% to 83.5% depending on VOCs composition that can transition from VOC-limited conditions to NO x -limited for O 3 formation. Using satellite measurements of HCHO and NO 2 , Jin et al. [15] estimated that transition from VOCs-limited to NO x -limited conditions occurred within 40-60 km for NYC by 2013-2016, as compared to 80-120 km in the past, accompanied by the reversal of O 3 weekend effect.
The transition from VOCs-limited to NO x -limited conditions may be better delineated in O 3 weekend effect. The average OWE effect for the urban areas was consistent with that observed in other US urban areas [31]. It has been attributed to the reduction in NO x emissions from road traffic on weekends, particularly on Sundays, leading to a lower O 3 titration by NO that also appears to be the dominant cause in NYC. VOCs emissions from recreational and residential activities may offset reduced traffic-related VOCs emissions in weekends allowing for longer O 3 accumulation and production [11]. In peri-urban sites, there was no O 3 weekend effect, in agreement with previous studies [30,31]. This was ascribed to the reduced NO titration throughout the week. Analysis of the 2018 summer O 3 exceedances in NYC, accompanied by a series of heatwaves showed that shortly downwind of Manhattan and within the urban corridor, O 3 formation transitioned to NO x -limited conditions. Moreover, the combination of NO x and biogenic VOCs primarily contributed to high O 3 levels [32].
We observed a strong correlation between O 3 levels and the frequency of regional wildfires. Changes in local photochemistry and regional transport may also influence O 3 trends. For the Northeastern US, ambient ozone concentrations were more dependent on ambient temperature (30%) than anthropogenic NO emissions reductions (10%) [33]. For the Northeast, which includes the study area, regional O 3 transport (60%) explained most of the O 3 variability [33]. Transport from wildfires can modify O 3 at receptor sites. It was previously observed that the 8-h max O 3 concentration increased as the fire intensity increased due to the mixing of VOCs-rich wildfire plumes with NO x [11,18,24]. These conditions may further enhance the NO x limited conditions in NYC yielding high O 3 concentrations. Moreover, oxygenated VOCs released during wildfires (e.g., methoxy phenols) may react with NO x to form stable peroxyacetyl nitrates (PANs) that permanently remove NO x and moderate downwind O 3 levels [24,32]. The transition from El Niño to La Niña conditions over periods of two-three years of El Niño-Southern Oscillation (ENSO) is associated with increased wildfires in the US. This may be due to the accumulation of fresh biomass during the El Niño events including invasive grasses that trigger faster wildfire progression. During La Niña conditions in the following years, increased temperatures, reduced precipitation, and drought create conditions that promote fast-spreading wildfires [34].

Conclusions
The analysis showed increasing O 3 concentrations in sites within urban agglomerations while O 3 concentrations peri-urban have been declining. This was tentatively assigned to changes in the photochemical regime from VOC-limited to NO x -limited conditions. The weekend-weekday O 3 pattern indicated that reduced O 3 titration by NO has been declining increasing the nighttime O 3 carryover and promoting longer O 3 accumulation periods. Moreover, a strong correlation of O 3 levels with regional wildfires was computed. This was attributed to increased VOCs emissions and the formation of PANs in the smoke plume during transport and its mixing with ground-level air that can further augment NO x -limited conditions. The frequency and magnitude of wildfires in the eastern US were related to the sequence of El-Nino and La Nina events, with more lightning-ignited fires during dry periods. To mitigate increasing O 3 levels in densely populated areas, future emission control strategies should also consider the compounding global and regional effects of climate change.
Author Contributions: Conceptualization, I.G.K.; methodology, S.S. and I.G.K.; data curation, formal analysis, S.S.; writing-original draft preparation, S.S.; writing-review and editing, I.G.K.; supervision, I.G.K. All authors have read and agreed to the published version of the manuscript.

Conflicts of Interest:
The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.