Impacts of Regional Transport and Meteorology on Ground-Level Ozone in Windsor, Canada

: This study investigated impacts of regional transport and meteorology on ground-level ozone (O 3 ) in the smog season (April–September) during 1996–2015 in Windsor, Ontario, Canada. Data from ﬁve upwind stations in the US, which are within 310 km (i.e., Allen Park and Lansing in Michigan, Erie, National Trail School, and Delaware in Ohio), were included to assess the regional characteristics of O 3 . The ﬁve US stations showed high degrees of similarity with O 3 concentrations in Windsor, with overall strong correlations (r = 0.567–0.876 for hourly O 3 and r = 0.587–0.92 for 8 h max O 3 concentrations) and a low degree of divergence, indicating that O 3 pollution in the study area shares regional characteristics. Meteorological conditions played important roles in O 3 levels in Windsor. High O 3 concentrations were associated with southerly and southwesterly air mass from which polluted and hot air mass was transported and that enhanced local photochemical O 3 production. In contrast, northerly ﬂows brought in clean, cool, and dry air mass, and led to low O 3 concentrations. Strong correlations were found between numbers of days with 8 h max O 3 concentrations greater than 70 ppb and numbers of days with daily max temperature greater than 30 ◦ C, as well as between daily max temperatures and 8 h max O 3 concentrations. Nearly half (45%) of the high O 3 days ( ≥ 90th percentile) occurred in dry tropical weather during 1996–2015, and the 90th percentile 8 h max O 3 was associated with dry tropical weather. Occurrences of both southerly ﬂow hours and dry tropical weather type in the smog season increased during the study period. If there were more hot and dry days in the next few decades due to climate change, the e ﬀ ect of emission control on reducing peak O 3 values would be diminished. Therefore, continued regional and international e ﬀ orts are essential to control precursors’ emissions and to mitigate O 3 pollution in Windsor. the impacts of regional and local / synoptic meteorological conditions on smog season O 3 concentrations in Windsor during the 20-year study of


Introduction
Ground-level ozone (O 3 ) is a secondary air pollutant, produced by photochemical reactions between volatile organic compounds (VOC) and nitrogen oxides (NO X ) [1]. Both local emissions of O 3 precursors and regional transport of O 3 and its precursors have strong impacts on O 3 levels [1]. O 3 production is non-linearly related to levels VOCs and NO X , and mainly depends on the ratio of the two [2,3]. In urban areas where NO X levels are normally high, O 3 levels increase with VOC levels but decrease due to the nitric oxide (NO) scavenging effect. At low NO X levels, O 3 production is limited by NO X levels but with little impact by VOCs. The degree of O 3 scavenging by NO is strongly affected by the ratio of NO to NO 2 . With the same value of NO X (i.e., NO + NO 2 ), lower NO to NO 2 ratios lead to less O 3 being consumed [4]. The decreasing ratios of NO to NO 2 have been observed in recent Figure 1. O3 monitoring at Windsor (red star) and five USA sites (black star) and a weather station in Toledo (blue circle) (base maps adapted from Google Maps, coordinates of monitoring sites from the Ontario Ministry of the Environment, Conservation and Parks (MECP) [27] and United States Environmental Protection Agency (USEPA) [28]).
The prevailing wind direction in Windsor during the 20-year study period was from the southwest ( Figure S1). To investigate the regional characteristics of O3, five US O3 monitoring stations located in a range of 17 to 310 km away from Windsor were selected ( Figure 1). Two stations (i.e., Allen Park and Lansing) are in Michigan, and the other three stations (i.e., Erie, National Trail School or NTS, and Delaware) are in Ohio. The parameters of each selected station are listed in Table 1.   [27] and United States Environmental Protection Agency (USEPA) [28]).
The prevailing wind direction in Windsor during the 20-year study period was from the southwest ( Figure S1). To investigate the regional characteristics of O 3 , five US O 3 monitoring stations located in a range of 17 to 310 km away from Windsor were selected ( Figure 1). Two stations (i.e., Allen Park and Lansing) are in Michigan, and the other three stations (i.e., Erie, National Trail School or NTS, and Delaware) are in Ohio. The parameters of each selected station are listed in Table 1.  [28]. From the same website, the daily maximum 8 h concentrations for a given calendar day, i.e., the highest of the 24 possible 8 h average concentrations in that day (hereafter referred to as "8 h max O 3 ") [31] were downloaded for the five US stations during 1996-2015. Hourly meteorological parameters, including temperature, relative humidity, wind direction, wind speed, visibility, atmospheric pressure, and weather conditions (e.g., cloudy, snowy, and clear) at Windsor Airport (8 km southeast from the Windsor Downtown station) during 1996-2015 were downloaded from Environment and Climate Change Canada website [32]. The daily Spatial Synoptic Classification (SSC) data were downloaded from Kent State University website [33]. The SSC method classifies the weather conditions at a given location on a daily basis using surface-based observations, including temperature, dew point, cloud cover, atmospheric pressure, as well as south-north and west-east wind components. The SSC data are available in Windsor and Toledo ( Figure 1) during the 20-year study period [33]. Toledo was selected to represent large scale synoptic weather conditions in the study region.
The smog season 8 h max O 3 concentrations in Windsor were calculated with hourly O 3 concentrations following USEPA's "Guideline on data handling conventions for the 8 h ozone NAAQS (National Ambient Air Quality Standards)" [31].

Impact of Airmass Movement on O 3 Levels in Windsor
Backward trajectory was used to identify the geographical origins and pathways of air mass arrived in Windsor. Twenty-four-hour backward trajectories in Windsor in each day of the smog season during 1996-2015 were simulated with HYSPLIT [6,7]. The meteorological data are archived Eta Data Assimilation System (EDAS) files with a horizontal resolution of 80 km during 1997-March 2004 and EDAS 40 km during April 2004-2015 [34]. The start time of each trajectory was 15:00 EDT (19:00 UTC) when maximum hourly O 3 concentrations normally occurred in a day. The start height of trajectory was 500 m above the ground level to reflect half of the estimated summertime mixing height.
Each backward trajectory was overlaid on a compass with 36-bins in 10 • intervals to determine the direction from which the air mass arrived in Windsor. If a 24 h trajectory passed through several 10-degree sectors, only the last segment (18 h-24 h) of the trajectory was used to determine the direction of the air mass. Eight-hour max O 3 concentrations by each of the 36-direction of trajectories in smog season were calculated. Hierarchical clustering analysis [35] was used to classify trajectories into three clusters based on the 8 h max O 3 concentrations. Meteorological parameters by each cluster were also calculated to investigate the association between 8 h max O 3 concentrations and meteorological conditions.

Impact of Meteorological Parameters on O 3 Levels in Windsor
Both local and large-scale synoptic meteorological parameters were used to investigate their impacts on O 3 concentrations in Windsor. Spearman's rank correlation coefficients between hourly O 3 concentrations and each meteorological parameter (temperature, relative humidity, visibility, wind speed, and atmospheric pressure) in Windsor were calculated in smog season during the study period. cloudy, and rainy days) in the smog season were calculated to investigate the association between 8 h max O 3 concentrations and the weather conditions using ANOVA.
SSC was used to investigate the associations in Windsor between 8 h max O 3 concentrations in the smog season (April-September) and seven types of weather conditions, dry polar (DP), dry moderate (DM), dry tropical (DT), moist polar (MP), moist moderate (MM), moist tropical (MT), and transitional (TR), as outlined in Jing et al. [21]. Percentage of high O 3 days (8 h max O 3 >90th percentile) and 8 h max O 3 concentrations by each SSC weather pattern were calculated.

Similarity Analysis
Analysis of similarity among Windsor and the five USA sties was conducted to investigate the regional characteristics of O 3 during the study period. The following two methods were employed:  (1): where p is the total number of paired measurements, and C ij and C ik are the measured concentrations at the j (reference) and k (comparison) sites on the i-th day, respectively.

Directional O 3 Concentrations
Smog season mean O 3 concentrations in Windsor by wind direction are shown in Figure S2. Higher O 3 concentrations (39.5 ppb) was associated with air mass from the south and southwest (140 • -220 • ), where several industrial states of the USA are located (e.g., Michigan and Ohio), suggesting regional transport of polluted air mass. Lower O 3 concentrations (26 ppb) were associated with winds from the north (320 • -50 • ), suggesting clean air mass from northern Ontario.

Directional O3 Concentrations
Smog season mean O3 concentrations in Windsor by wind direction are shown in Figure S2. Higher O3 concentrations (39.5 ppb) was associated with air mass from the south and southwest (140°-220°), where several industrial states of the USA are located (e.g., Michigan and Ohio), suggesting regional transport of polluted air mass. Lower O3 concentrations (26 ppb) were associated with winds from the north (320°-50°), suggesting clean air mass from northern Ontario. Figure 2 depicts distribution of smog season O3 levels in Windsor in each wind direction. Under clean conditions (5th and 25th percentiles), the west and northwesterly flows (230°-360°) are associated with extremely low concentrations in comparison with other directions (2.7 ppb vs. 7.7 ppb for the 5th percentile; 16 ppb vs. 23 ppb for the 25th percentile). For median (50th) and high (75th and 95th) O3 concentrations, southerly flows (140°-220°) bring in significantly more polluted airmass than by flows from the north (330°-20°) (39 ppb vs. 23 ppb for the 50th percentile; 53 ppb vs. 39 ppb for the 75th percentile; 74 ppb vs. 47 ppb for the 95th percentile). Our results further indicate the strong impact of US states on higher O3 levels in Windsor. As a secondary pollutant, O3 levels depend strongly on regional transport of O3 and precursor concentrations. A study in Texas (USA) investigated impacts of cold front (mainly northerly winds) on area-wide peak levels and regional background concentrations during O3 seasons (April-October) of 2003-2016 [39]. They found that wind direction was the dominant factor causing changes on O3 levels, especially when southerly flow with less anthropogenic emissions from the Gulf of Mexico shifted to northerly flow with more emissions from inland. Plocoste et al. [40] reported strong influences of high O3 levels by meteorological conditions (e.g., wind) in Guadeloupe, a French overseas region located at the Lesser Antilles Arc. The impacts of regional transport from upwind states on high O3 levels have also been reported by other researchers, for example, in the states of New York [8] and Delaware [20].
A USEPA study used the Variable Grid Urban Airshed Model (UAM-V) to determine the pattern of transport in the transboundary region of the eastern United States and Canada [41]. It was found that the magnitude and persistence of high O3 concentrations in eastern North America are strongly influenced by long-range transport of O3 and its precursors. As for transboundary movement of pollutants, more O3 and precursors transport from the US into Canada than from Canada into the USA. Consequently, air mass from the US contributed to high O3 concentrations in Ontario, Quebec, As a secondary pollutant, O 3 levels depend strongly on regional transport of O 3 and precursor concentrations. A study in Texas (USA) investigated impacts of cold front (mainly northerly winds) on area-wide peak levels and regional background concentrations during O 3 seasons (April-October) of 2003-2016 [39]. They found that wind direction was the dominant factor causing changes on O 3 levels, especially when southerly flow with less anthropogenic emissions from the Gulf of Mexico shifted to northerly flow with more emissions from inland. Plocoste et al. [40] reported strong influences of high O 3 levels by meteorological conditions (e.g., wind) in Guadeloupe, a French overseas region located at the Lesser Antilles Arc. The impacts of regional transport from upwind states on high O 3 levels have also been reported by other researchers, for example, in the states of New York [8] and Delaware [20].
A USEPA study used the Variable Grid Urban Airshed Model (UAM-V) to determine the pattern of transport in the transboundary region of the eastern United States and Canada [41]. It was found that the magnitude and persistence of high O 3 concentrations in eastern North America are strongly influenced by long-range transport of O 3 and its precursors. As for transboundary movement of pollutants, more O 3 and precursors transport from the US into Canada than from Canada into the USA. Consequently, air mass from the US contributed to high O 3 concentrations in Ontario, Quebec, New Brunswick and Nova Scotia, Canada. This is because of higher emissions in the USA and the prevailing winds during the summer O 3 season. Commission for Environmental Cooperation reported significant transport of pollutants into southwest Ontario from the upper Midwest US States and the Ohio River Valley [5].
Long-term trend of southerly flow (160 • -200 • ) hours in the smog season in Windsor is shown in two lowest annual means (25.5 ppb and 26.6 ppb, respectively), but high in southerly flow hours (638 h and 774 h, respectively) during 1996-2015. The remaining 18 years showed a strong correlation (r = 0.642, p < 0.05). Therefore, increasing southerly flows could be one of the reasons of increasing O 3 concentrations in smog season in Windsor reported in Zhang. [30].
prevailing winds during the summer O3 season. Commission for Environmental Cooperation reported significant transport of pollutants into southwest Ontario from the upper Midwest US States and the Ohio River Valley [5].
Long-term trend of southerly flow (160°-200°) hours in the smog season in Windsor is shown in Figure 3. The occurrence of southerly flow hours increased in Windsor by 30% (p < 0.05) during the 20-year study period. Furthermore, annual smog season hours of southerly flow were positively correlated with annual smog season O3. The two exceptions are year 2000 and year 2004 with the two lowest annual means (25.5 ppb and 26.6 ppb, respectively), but high in southerly flow hours (638 h and 774 h, respectively) during 1996-2015. The remaining 18 years showed a strong correlation (r = 0.642, p < 0.05). Therefore, increasing southerly flows could be one of the reasons of increasing O3 concentrations in smog season in Windsor reported in Zhang. [30].

Air Mass Trajectory
Smog season daily 24 h HYSPLIT backward trajectories ending in Windsor during the study period are organized by year in Google Earth format. An example of a backward trajectory plot in 2015 is shown in Figure S3. Backward trajectory plots shared a common feature in each year of the study period, with lower O3 (first quartile, green lines) and higher O3 (forth quartile, red lines) mostly from the north and south, respectively, while the middle levels (second quartile, yellow; third quartile, pink) were from all directions. This once again indicated the impact of regional transport of polluted air mass from the south on higher O3 levels in Windsor.
Smog season 8 h max O3 concentrations averaged by 36 air mass trajectory directions arriving in Windsor during the 20-year study period are shown in Figure S4, the trend is similar to that of the directional concentrations by wind direction ( Figure S2) using hourly data. Those 36 directional air trajectories were classified into three clusters. Eight-hour max O3 concentrations by cluster and meteorological conditions of each cluster are summarized in Table 3. One-way ANOVA indicated that there are at least two cluster means of 8 h max O3 concentrations and of each meteorological parameter considered were significantly different from each other (Table S1).

Air Mass Trajectory
Smog season daily 24 h HYSPLIT backward trajectories ending in Windsor during the study period are organized by year in Google Earth format. An example of a backward trajectory plot in 2015 is shown in Figure S3. Backward trajectory plots shared a common feature in each year of the study period, with lower O 3 (first quartile, green lines) and higher O 3 (forth quartile, red lines) mostly from the north and south, respectively, while the middle levels (second quartile, yellow; third quartile, pink) were from all directions. This once again indicated the impact of regional transport of polluted air mass from the south on higher O 3 levels in Windsor.
Smog season 8 h max O 3 concentrations averaged by 36 air mass trajectory directions arriving in Windsor during the 20-year study period are shown in Figure S4, the trend is similar to that of the directional concentrations by wind direction ( Figure S2) using hourly data. Those 36 directional air trajectories were classified into three clusters. Eight-hour max O 3 concentrations by cluster and meteorological conditions of each cluster are summarized in Table 3. One-way ANOVA indicated that there are at least two cluster means of 8 h max O 3 concentrations and of each meteorological parameter considered were significantly different from each other (Table S1). Median level 8 h max O 3 concentrations (46 ppb) in Windsor were associated with air mass from the east and southeast (60 • to 150 • , Cluster 2), including Pennsylvania, New York, and Eastern Ohio. Air mass in this cluster is the least frequent (18%) and associated with cool, humid, and slow-moving conditions. The highest percentage of rainy days at 9.3% was found in Cluster 2, suggesting unfavorable conditions for O 3 formation.
Eight-hour max O 3 concentrations in Cluster 3 were the highest (51 ppb) when air mass was from the south and southwest (160 • to 290 • ) of Windsor, where several industrial states are located including Michigan, western Ohio, Indiana, and southern Illinois. Air mass from Cluster 3 is the most frequent (50%), consistent with prevalent wind direction in Windsor ( Figure S1). Air mass in this cluster led to hot days with strong winds. The high temperatures favor photochemical production of O 3 . Our results suggest that high O 3 concentrations in Windsor were caused by both local photochemical production under strong solar radiation and regional transport of O 3 and its precursors from the industrial states of the US. Similar findings were reported by other researchers. A study of O 3 concentrations in Eastern Canada founded that the elevated levels of warm season O 3 are associated with back-trajectories originating from Detroit-Windsor or the Ohio River Valley region [42]. Another O 3 modeling study [43] in Southern Ontario in the summer of 2001 reported that around 60% of the O 3 formed by anthropogenic emissions in Southern Ontario was due to the emission releases from the nearby US states during the smog episodes.
Among the four meteorological parameters considered in Table 3, temperature and atmospheric pressure are the only two parameters that distinguish Cluster 3 from the other two clusters, while low temperature and relative humidity are unique to Cluster 1 (Table S1). This finding suggests that during the smog season, when hot air mass come from the south and southwest of Windsor, O 3 concentration is likely high; while cool and dry air mass from the north tends to lead to lower O 3 levels in Windsor.

Effects of Meteorological Parameters
Spearman's rank correlation coefficients and Pearson correlation coefficients between smog season O 3 concentrations and five continuous meteorological parameters in Windsor are provided in Table S2. Of the five meteorological parameters investigated, only temperature (hourly data, r = 0.518, p < 0.01) and relative humidity (hourly data, r = −0.521, p < 0.01) were strongly correlated with O 3 concentrations during 1996-2015. This suggests faster photochemical production of O 3 under warmer and drier conditions [44].
As depicted in Figure 4, the numbers of days of smog season 8 h max O 3 greater than 70 ppb was strongly correlated with the numbers of hot days (i.e., daily max temperature greater than 30 • C) in Windsor during the study period (r = 0.833, p < 0.05). Furthermore, there was a strong correlation between daily max temperature and 8 h max O 3 concentrations (r = 0.714, p < 0.05) in smog season. This O 3 concentration-temperature association is consistent with findings from an O 3 study in the Greater Toronto Area of Ontario [17]. Association between elevated O 3 levels and high temperatures has also been reported by other studies [3,23,24,40].
Atmosphere 2020, 11, x FOR PEER REVIEW 10 of 17 This O3 concentration-temperature association is consistent with findings from an O3 study in the Greater Toronto Area of Ontario [17]. Association between elevated O3 levels and high temperatures has also been reported by other studies [3,23,24,40]. Mean meteorological parameters in high (>90th) and low (<10th) smog season 8 h max O3 concentration days are summarized in Table 4. One-way ANOVA and Tukey's tests indicate that all meteorological parameters were significantly different (p < 0.05) in high and low O3 days, indicating strong influence by weather conditions. High O3 days (>90th percentile 8 h max O3) were typically associated with high temperature and atmospheric pressure, but low in relative humidity and wind speed. Hot and dry days are generally associated with strong solar intensity, thus enhancing photochemical production of O3 [45]. High pressure and low wind speeds weaken the dispersion and dilution of O3, resulting in high O3 concentrations. The association between light winds and elevated O3 levels was also reported by Johnson et al. [42] in Southern Ontario and Southwest Quebec (Canada) during 1994-2003, and by Davis et al. [12] in  Mean meteorological parameters in high (>90th) and low (<10th) smog season 8 h max O 3 concentration days are summarized in Table 4. One-way ANOVA and Tukey's tests indicate that all meteorological parameters were significantly different (p < 0.05) in high and low O 3 days, indicating strong influence by weather conditions. High O 3 days (>90th percentile 8 h max O 3 ) were typically associated with high temperature and atmospheric pressure, but low in relative humidity and wind speed. Hot and dry days are generally associated with strong solar intensity, thus enhancing photochemical production of O 3 [45]. High pressure and low wind speeds weaken the dispersion and dilution of O 3 , resulting in high O 3 concentrations. The association between light winds and elevated O 3 levels was also reported by Johnson et al. [42] in Southern Ontario and Southwest Quebec (Canada) during 1994-2003, and by Davis et al. [12] in Shenandoah Valley of Virginia (USA) during 2001-2006.
Low O 3 days (<10th percentile 8 h max O 3 ) were usually accompanied by low temperature and atmospheric pressure, but humid and windy weather. Cooler and humid days are associated with weak solar intensity, thus result in less photochemical production of O 3 [45]. Low atmospheric pressure and strong winds enhance dispersion and dilution of O 3 , leading to low O 3 concentrations.
One-way ANOVA also indicates that smog season 8 h max O 3 concentrations in clear days (51 ppb) were statistically higher (p < 0.05) than those in cloudy days (45 ppb) and rainy days (33 ppb), which are unfavorable to O 3 formation.

Impact of Synoptic Weather Types
The occurrence frequency of SSC weather types during smog season was calculated for every 5-year period from 1996 to 2015 to assess how it had changed over time, as shown in Figure 5. Warmer weather types (DM, DT, MM, and MT) dominated the study region in comparison with cooler (DP and MP) or transitional (TR) weather types.
Atmosphere 2020, 11, x FOR PEER REVIEW 11 of 17 Low O3 days (<10th percentile 8 h max O3) were usually accompanied by low temperature and atmospheric pressure, but humid and windy weather. Cooler and humid days are associated with weak solar intensity, thus result in less photochemical production of O3 [45]. Low atmospheric pressure and strong winds enhance dispersion and dilution of O3, leading to low O3 concentrations.
One-way ANOVA also indicates that smog season 8 h max O3 concentrations in clear days (51 ppb) were statistically higher (p < 0.05) than those in cloudy days (45 ppb) and rainy days (33 ppb), which are unfavorable to O3 formation.

Impact of Synoptic Weather Types
The occurrence frequency of SSC weather types during smog season was calculated for every 5year period from 1996 to 2015 to assess how it had changed over time, as shown in Figure 5. Warmer weather types (DM, DT, MM, and MT) dominated the study region in comparison with cooler (DP and MP) or transitional (TR) weather types. The number of high O3 days (8 h max O3 > 90th percentile, shown as solid blue bars in Figure 5) were mainly associated with warm and dry weather types (i.e., DM, DT, and MT), and rarely observed in cool weather types (i.e., DP and MP), high cloudy (MM) or transitional days (TR). The largest number of high O3 days (averaging 8.5 days/yr) was observed under DM weather, primarily because Windsor was influenced by DM weather more often than by other weather types in the smog season. Although DT days were far less frequent than DM and MT days, a high number of high O3 days (7.8 days/yr) was observed under DT weather type. In the past 20 years, high O3 concentrations occurred in nearly half of DT days (45%), much higher than those under the other six weather types (MT 16%, DM 13%, TR 2%, MM 0.6%, DP 0%, and MP 0%). In general, the number of high O3 days decreased or unchanged in all weather types, except for a steady increase in DT weather type.
Trends of DT days and 90th percentile 8 h max O3 during the past 20 years are shown in Figure  6. There was a slight increase in the number of DT days and a marginal decrease in the 90th percentile 8 h max O3 concentration. There was a clear association between these two parameters (r = 0.644, p < The number of high O 3 days (8 h max O 3 > 90th percentile, shown as solid blue bars in Figure 5) were mainly associated with warm and dry weather types (i.e., DM, DT, and MT), and rarely observed in cool weather types (i.e., DP and MP), high cloudy (MM) or transitional days (TR). The largest number of high O 3 days (averaging 8.5 days/yr) was observed under DM weather, primarily because Windsor was influenced by DM weather more often than by other weather types in the smog season. Although DT days were far less frequent than DM and MT days, a high number of high O 3 days (7.8 days/yr) was observed under DT weather type. In the past 20 years, high O 3 concentrations occurred in nearly half of DT days (45%), much higher than those under the other six weather types (MT 16%, DM 13%, TR 2%, MM 0.6%, DP 0%, and MP 0%). In general, the number of high O 3 days decreased or unchanged in all weather types, except for a steady increase in DT weather type.
Trends of DT days and 90th percentile 8 h max O 3 during the past 20 years are shown in Figure 6. There was a slight increase in the number of DT days and a marginal decrease in the 90th percentile 8 h max O 3 concentration. There was a clear association between these two parameters (r = 0.644, p < 0.05), i.e., annual peak 8 h max O 3 values were higher when there were more DT days in that year. If there were more DT days in the next few decades in the study region due to climate change [21], peak O 3 concentrations would be increased and the reduction in peak O 3 concentrations as the result of emission controls would be diminished.
Atmosphere 2020, 11, x FOR PEER REVIEW 12 of 17 0.05), i.e., annual peak 8 h max O3 values were higher when there were more DT days in that year. If there were more DT days in the next few decades in the study region due to climate change [21], peak O3 concentrations would be increased and the reduction in peak O3 concentrations as the result of emission controls would be diminished.

Similarity in O3 Concentrations between Windsor and the Five US Sites
3.3.1. Correlation Table 5 lists correlation coefficients between O3 concentrations in Windsor and at each of the five USA sites during 1996-2015. Strong correlations (r > 0.74) between O3 at Windsor and Allen Park, Erie, and Lansing sites indicate a good agreement in temporal variability of O3 concentrations. Moderate correlation (0.72 > r > 0.4) between O3 at Windsor and NTS as well as Delaware sites suggest less similarity with those two areas which are rural sites and further away from the Windsor ( Figure  1). The results of correlation analysis suggest regional characteristics of O3 pollution in the study area. Likewise, correlation coefficients of 8 h max O3 concentrations in the greater Chicago areas (41.3° N-42.6° N, 87.0° W-88.5° W) were calculated during 2005-2013 in the summer months of May to August [38]. The correlation coefficients of 8 h max O3 between any two sites ranged 0.71-0.94, suggesting a similar temporal variability of O3 among the 23 monitoring stations. The lower correlation coefficients were observed at those sites that are furthest apart in the north-south direction.  [38]. The correlation coefficients of 8 h max O 3 between any two sites ranged 0.71-0.94, suggesting a similar temporal variability of O 3 among the 23 monitoring stations. The lower correlation coefficients were observed at those sites that are furthest apart in the north-south direction.
By examining 8 h max O 3 concentrations by cluster, considerately higher correlation coefficients between Windsor and the five USA sites were observed in Cluster 3 (high O 3 ) than in Clusters 1 and 2 (Table 6). This indicates that higher O 3 levels associated with southerly and southwesterly flow exhibited a stronger regional signal.   Table 3). All correlations are significant at p < 0.05.

Divergence
As shown in Table 7, the COD values for 8 h max O 3 concentrations (0.099-0.148) were lower than those of hourly O 3 (0.260-0.338) due to greater spatial variability in the latter. However, COD rankings of O 3 concentrations between Windsor and each of the five US sites were consistent in hourly and 8 h max concentrations. O 3 concentrations between Windsor and Lansing exhibited the most homogeneity, followed by Allen Park and Erie. Delaware and NTS had less homogeneity with Windsor. This result is largely consistent with the rankings by correlation coefficients (Tables 5 and 6), once again indicating regional characteristics of O 3 in the study area. Other researchers also reported comparable COD values among sites with similar characteristics. A study in Treasure Valley (Idaho, US) estimated COD of O 3 concentrations between a reference site (PAR) and six other sites during 1 July to 30 September 2007 [37]. Lower COD values were observed between PAR and two sites (0.06) near downtown Boise, higher COD values between PAR and two sites (0.11-0.17) on the southeast end of the valley, and moderate COD between PAR and two sites (0.08-0.09) located upwind of Boise and influenced by significant mobile sources.

Conclusions
This study assessed impacts of regional transport and meteorology on O 3 concentrations in Windsor, Canada, by using data collected from 1996-2015. O 3 concentrations in the smog season exhibited a high degree of similarity between Windsor and the five USA sites within 310 km, suggesting that the O 3 pollution in Windsor and at the five sites in the US showed regional characteristics, likely due to their similar emission sources of O 3 precursors and the shared weather conditions. High O 3 concentrations in Windsor were found to be associated with southerly and southwesterly flows that brought in hot and polluted air mass enhancing local photochemical production. In contrast, northerly flows brought in clean, dry and cool air mass from northern Ontario, leading to lower O 3 concentrations. Therefore, regional transport of O 3 and its precursors from upwind areas had a great impact on O 3 levels in Windsor.
Strong correlations were found between numbers of days with 8 h max O 3 concentrations greater than 70 ppb and the numbers of days with daily max temperature greater than 30 • C, as well as between daily max temperatures and 8 h max O 3 concentrations. The peak O 3 (8 h max O 3 > 90th percentile) days were associated with hot and calm weather conditions, characterized by high temperature and atmospheric pressure, but low relative humidity and wind speed, belonging to dry tropical weather type, i.e., the hottest and driest synoptic weather conditions. This study showed that both regional transport and local photochemical production played an important role in the peak O 3 levels in smog season in Windsor during the past 20 years. Occurrences of both southerly flow hours and dry tropical weather type in the smog season, which were associated with high 8 h max O 3 concentrations, increased during the study period. If there were more hot and dry days in the next few decades due to climate change [3], the effect of emission control on reducing peak O 3 values would be diminished. Given increased DT days in the Windsor area, continued regional and international efforts are essential to control precursors' emissions and to mitigate high O 3 levels.  Table S1. Result of Tukey's tests of 8-h max O3 and meteorological parameters by cluster in Windsor during 1996-2015 (in Table 3). Table S2. Correlation coefficients between smog season hourly/daily O3 concentrations and meteorological parameters in Windsor during 1996-2015.
Author Contributions: Data collection, data, analysis, visualization, manuscript draft preparation, T.Z.; conceptualization, review and editing of manuscript, X.X. and Y.S. All authors have read and agreed to the published version of the manuscript.