**Stratosphere–Troposphere Exchange and O3 Variability in the Lower Stratosphere and Upper Troposphere over the Irene SHADOZ Site, South Africa**

#### **Thumeka Mkololo 1,2,\*, Nkanyiso Mbatha 3, Venkataraman Sivakumar <sup>1</sup> , Nelson Bègue 4, Gerrie Coetzee <sup>2</sup> and Casper Labuschagne <sup>2</sup>**


Received: 17 February 2020; Accepted: 7 May 2020; Published: 3 June 2020

**Abstract:** This study aims to investigate the Stratosphere-Troposphere Exchange (STE) events and ozone changes over Irene (25.5◦ S, 28.1◦ E). Twelve years of ozonesondes data (2000–2007, 2012–2015) from Irene station operating in the framework of the Southern Hemisphere Additional Ozonesodes (SHADOZ) was used to study the troposphere (0–16 km) and stratosphere (17–28 km) ozone (O3) vertical profiles. Ozone profiles were grouped into three categories (2000–2003, 2004–2007 and 2012–2015) and average composites were calculated for each category. Fifteen O3 enhancement events were identified over the study period. These events were observed in all seasons (one event in summer, four events in autumn, five events in winter and five events in spring); however, they predominantly occur in winter and spring. The STE events presented here are observed to be influenced by the Southern Hemisphere polar vortex. To strengthen the investigation into STE events, advected potential vorticity maps were used, which were assimilated using Modélisation Isentrope du transport Méso–échelle de l'Ozone Stratosphérique par Advection (MIMOSA) model for the 350 K (~12–13 km) isentropic level. These maps indicated transport of high latitude air masses responsible for the reduction of the O3 mole fractions at the lower stratosphere over Irene which coincides with the enhancement of ozone in the upper troposphere. In general, the stratosphere is dominated by higher Modern Retrospective Analysis for Research Application (MERRA-2) potential vorticity (PV) values compared to the troposphere. However, during the STE events, higher PV values from the stratosphere were observed to intrude the troposphere. Ozone decline was observed from 12 km to 24 km with the highest decline occurring from 14 km to 18 km. An average decrease of 6.0% and 9.1% was calculated from 12 to 24 km in 2004–2007 and 2012–2015 respectively, when compared with 2000–2003 average composite. The observed decline occurred in the upper troposphere and lower stratosphere with winter and spring showing more decline compared with summer and autumn.

**Keywords:** ozone enhancement; Irene; ozone decline; potential vorticity; ozonesondes

#### **1. Introduction**

The stratosphere and troposphere have different characteristics that are useful in identifying air movement from the stratosphere to the troposphere and vice versa. The stratosphere is characterized by high ozone (O3) and potential vorticity (PV). On the other hand, the troposphere is characterized by low O3 and PV. Approximately 90% of O3 is found in the stratosphere and only 10% in the troposphere. Most of the O3 in the stratosphere is situated within the O3 layer [1] where it plays a critical role in shielding the environment and protecting human health from dangerous ultraviolet (UV) rays. The O3 hole was discovered over Antarctica in late 1980's [2,3] and its dynamics with Ozone Depleting Substances (ODS) are well documented [4–6]. The appearance of O3 hole was a big concern because of the relationship that increased UV can have on various processes in the lower troposphere. In 1998, the Montreal protocol was successfully implemented to phase out the use of ODS. Subsequently, ozone was expected to increase in the stratosphere after ODS were phased out. Since 1985, a decreasing trend of 30 DU in total column O3 was reported for stations over the southern mid-latitudes [7]. A positive trend was only observed in the upper stratosphere above 10 hPa [8,9]. A continuous O3 decline (even though statistically non-significant) was reported in the lower stratosphere from 1998 until present for the stations lying between 60◦ N and 60◦ S [10].

Tropospheric O3 mole fractions are controlled by chemical and physical processes, and its precursors originate from both natural and anthropogenic activities. In the troposphere, O3 is a secondary pollutant formed through a number of reactions containing nitrogen oxides (NOx), volatile organic compounds (VOCs), methane (CH4), and CO in the presence of sunlight [11] or through the transportation of O3–rich air from the stratosphere. This process of O3 movement from the stratosphere to the troposphere and vice versa is known as stratosphere–troposphere O3 exchange (STE). After the Los Angeles photochemical smog, studies revealed that O3 can either be transported from the stratosphere or produced from chemical reactions involving O3 precursors in the troposphere [12]. Stratosphere intrusion (SI) are expected to transport high O3 and PV to the troposphere. Hence, high O3, high PV and dry atmospheric air are used to identify stratosphere intrusion events in the troposphere. The STE plays an important role in the chemical budget of O3 and water vapour of the upper troposphere and lower stratosphere [13]. Stratospheric Intrusion studies are poorly documented in the Southern Hemisphere; hence, it is challenging to find the threshold that could be used to study these events in the literature. Ozone vertical profiles are useful in identifying and studying STE. A number of studies have been undertaken using Irene ozonesondes data. These studies include: (a) study of O3 climatology over Irene by Diab et al. [14] using 1990 to 1994 and 1998 to 2004 dataset, (b) satellite validation [15,16] of global transport models, (c) a study on stratospheric profile and water vapour in Southern Hemisphere [17], and (d) the study on the Southern Hemisphere tropopause [18]. In their study, Diab et al. [14] reported O3 enhancement in Irene above 10 km (upper troposphere) which can be related to STEs during late winter. Similar to this study, they also noted the absence of seasonal consistency in the occurrence of these events at a height above 10 km level. The frequency of occurrence of STEs in the Irene SHADOZ data set has never been studied due to limited data availability and the frequency of the launching of ozonesondes. However, according to Diab et al. [14], the STEs events are dominant in winter and spring in the Irene station. In support to these observations, a study by Poulida et al. [19] also found high O3 mole fractions in the upper troposphere dominating in winter and spring months.

There are limited studies conducted over Southern Africa on STEs. A recent study on high O3 events was conducted by Mulumba in Nairobi, Congo Basin and Irene using ozonesondes data [20]. However, the focus was more on Nairobi and Congo Basin and little was done with Irene data. Hence, this study focuses on O3 data observed from Irene station. It is crucial to investigate O3 enhancement events in the troposphere and determine if such episodes were related to stratosphere–troposphere exchange. Also, events such as Cut–off lows, Rossby waves, Quasi-Biennial Oscillation (QBO) and El Niño-Southern Oscillation (ENSO) are all factors that could potentially play a critical role in STEs [21,22]. For example, QBO affects the troposphere by direct effect of QBO on the tropical or

subtropical troposphere [23]. A downward movement of easterly winds is more dominant and much stronger compared to westerlies. Easterlies are represented as negative on the QBO index while positive signal represents westerlies. Another way that QBO affects the troposphere is through polar vortex [24–26] processes. Most researchers have defined the polar vortex as a region of high PV. The stratosphere polar vortex develops in autumn when there is no solar heating in polar regions, strengthens in winter and breaks down in spring as sunlight returns to polar regions [27]. The breakdown of stratospheric polar vortex especially plays an important role in O3 distribution in high latitudes [22]. Several studies have investigated the role of the southern polar vortex with respect to the middle atmosphere of the southern hemisphere [28–30]

A comprehensive study on STEs was recently conducted at three Southern Hemisphere stations (Davis (69◦ S, 78◦ E), Macquarie Island (55◦ S, 159◦ E) and Melbourne (38◦ S, 145◦ E)) by looking at the statistical analysis of STEs and their impact on tropospheric O3 [31]. This study has coupled observed STEs with meteorological conditions such as low pressure fronts, cutoff low pressure system, indeterminate meteorology and smoke plumes. A total number of 45, 47 and 72 events were detected in Davis, Macquarie and Melbourne stations, respectively, from ozonesondes data. The majority of events were related to low pressure fronts with fire plumes contributing the least.

A number of researchers have investigated O3 trends in the lower troposphere O3 [32–36] and only a few studies have conducted trends analysis at different altitudes of troposphere and the lower stratosphere [9,37]. A study by Granados–Munoz and Leblanc [38] investigated tropospheric trends at different altitudes over California using a procedure similar to that described by Cooper et al. [33]. In their study, linear fits of medians, 5th percentiles and 95th percentiles were done using least squares method. Statistical significant negative trends were observed in the lower troposphere (4–7 km) in winter for the medians and 5th percentiles. On the other hand, a positive significant trend of 0.3 ppb/year was reported for the upper troposphere (7 to 10 km) for the period of 2000 to 2015. A nonsignificant trend was reported for layers closer to the tropopause whilst negative trends were observed in the lower stratosphere (17 to 19 km). A recent study by Ball [9] reported a continuous decrease of O3 in the lower stratosphere in the region between 60◦ N and 60◦ S, while ozone recovery was observed in the upper stratosphere.

The aim of this study is to utilise the available Irene ozonesonde vertical profile to investigate the STE events that are known to lead to SI occurrence. Thus, this study identifies O3 enhancement events exceeding monthly 90th percentile composite in the upper troposphere, and investigate whether such episodes were due to SI events. Another objective of this study is to investigate O3 changes in the upper troposphere and lower stratosphere using Irene ozonesondes data and linear regression for medians, 5th percentiles and 95th percentiles.

#### **2. Method and Data**

#### *2.1. Ozonesondes*

Irene station soundings started in 1998 during the Southern African Fire Atmospheric Research Initiative (SAFARI) campaign in the African region. Currently, the station operates within the framework of the Southern Hemisphere Additional Ozonesondes (SHADOZ). The main aim of the project was to determine O3 mole fractions in the troposphere and stratosphere, and also to have a full coverage of O3 measurements over Southern Hemisphere. Since then, balloon launching continued in Irene with ozonesondes launched every second Wednesday of each month circumstances allowing. Ozone vertical profiles are obtained using electrochemical concentration cell (ECC). The heart of the instrument is the electrochemical cell that interfaces with a radiosonde that transmits back data signals to the ground station receiver. The method used by the electrochemical cell to detect O3 was discussed extensively by Sivakumar et al. [39]. A total number of 250 ozonesondes were launched over the study period. Irene ozonesondes data was retrieved from SHADOZ website [40]. It is important to note that there is a data gap of approximately four years (2008 to September 2012) in the Irene data due to

budget constraints and technical problems with the ground receiver. The program was resumed in 2012 when these issues were resolved.

#### *2.2. MERRA-2 Potential Vorticity*

The stratosphere has static stability and is known to contain higher potential vorticity (PV) compared with the troposphere. During the stratosphere intrusion episode, an air mass rich in O3 and high PV enters the lower stratosphere and upper troposphere. Hence, PV can be used to identify troposphere air mass having a stratosphere origin. The tropopause is defined using a PV value as 2 PVU [39,41]. Therefore, any higher PV events located in the troposphere are associated with stratosphere origin. Other studies, have used PV values of 1.5 PVU as a threshold to identify stratosphere air [42,43]. In this study, a PV value of 2 PVU was used as a threshold for stratospheric air. This study employs PV data from the Modern Retrospective Analysis for Research Application version 2 (MERRA-2) with a spatial resolution of 0.5 × 0.625◦. MERRA-2 model is an Earth System reanalysis model by National Aeronautics Space Administration (NASA) Global Modeling and Assimilation Office (GMAO). More details about MERRA-2 can be found in the website: https://gmao.gsfc.nasa.gov/reanalysis/MERRA-2/. Although the units of PV are Km2 kg<sup>−</sup>1s−1, PV Units (PVU) (where 1 PVU = 1 <sup>×</sup> 10−<sup>6</sup> Km2 kg-1s−1) will be used for convenience in this study. This study used NASA instrument Panoply software (see https://www.giss.nasa.gov) to view the vertical slices of PV over Irene [44].

#### *2.3. Data Processing*

The ozonesonde data is recorded every two seconds from 1.5 km to approximately 28 km. Ozone averages were calculated from 2 seconds data for each kilometer (km) ascended (e.g., 1 km, 2 km, 3 km to 28 km). As the ozonesonde ascends, pressure, temperature, humidity and O3 (both ppm and DU units) are also recorded. The current study used the available ozonesondes data to investigate high O3 events. The ozonesonde data was grouped into months in order to calculate averages, standard deviations, medians, 5th and 95th percentiles from 1 to 28 km. Monthly averaged data was used in conjunction with individual high O3 event profiles to determine how individual profiles differ from their respective monthly composite profiles. This study defines high O3 event as the event where O3 exceeds the monthly 90th and 95th percentile composites. Events exceeding these percentile composites were selected as an observed high O3 events. However, these events might originate from different sources such as plumes, stratospheric-troposphere exchange (STE) and other man made activities that generate O3 precursors. Due to this reason, potential vorticity (PV) of 2PV was added as another criteria to identify events of stratospheric origin. PV was used in this study because it is one of the characteristics to differentiate between the stratosphere and troposphere air masses. Monthly 90th and 95th percentile composites were calculated from the available data of all ozonesondes launched during the study period. Any profile exceeding the monthly 90th and 95th percentile composites at a height between 6 and 11 km was considered for further investigation. The main reason to focus between 6 km and 11 km is to eliminate tropospheric pollution and to select events that occurred below the tropopause. Our method differs from previous studies that identified Stratospheric Intrusion (SI) events as an event where O3 exceeds 80 ppb and then within 3 km above decreases by 20 ppb or more to a value less than 120 ppb [45]. This method was not used because it will miss some of the events observed over Irene due to lower O3 mole fractions in the Southern Hemisphere. Another study used the 99th percentile as a threshold to study SI O3 events and their impact on tropospheric ozone [31]. Similar percentile threshold was not applied in this study for similar reason stated above. We attempted to use the 95th percentile as a threshold, however, we missed five events. Consequently, we opted to use 90th percentile as a threshold.

Average O3 composite profiles were calculated for three categories (2000–2003, 2004–2007 and 2012–2015) using the available ozonesondes data. In the case of annual and seasonal O3 changes, monthly averaged composites were calculated from the available ozonesondes data. Monthly averaged

composites were used to compensate data gaps that occurred in summer and autumn. Hence, only summer and autumn composites were used to fill 2015 data gaps. This exercise was done to prevent the bias that may be caused by months with data gaps in calculating O3 changes at different altitudes. Data gaps were covered only for 2015, not for the years where there was no ozonesondes data for the complete year. Annual changes were studied by averaging monthly data into yearly averages at different layers such as 13–15 km, 16–18 km, 19–21 km and 22–24 km. Furthermore, the medians, 5th and 95th percentiles were calculated at each layer. The slope was determined by fitting a linear trend to yearly averaged data plotted on the scatter plot. The standard error corresponding to the slope was calculated at each layer for median values, and both the 5th and 95th percentile. Similar approach was used to calculate seasonal O3 changes.

#### **3. Results and Discussion**

Figure 1 indicates the monthly (a) and yearly (b) ozonesondes data launched from year 2000 to year 2015 at Irene station. Over the study period, a total number of 250 ozonesondes were launched from January 2008 to September 2012 showing a significant data gap. However, 12.4% of the launched ozonesondes did not reach above 28 km. A maximum of 25 ozonesondes were launched in October and November, respectively whilst a total of 12 ozonesondes were launched in January over the study period. On a per annual basis, the highest number (39) of ozonesondes were launched in 2000 while 2015 reflects the lowest number (11) of ozonesondes launched. The discrepancy in the annual number of ozonesondes launched was a factor of budgetary constraints as well as some operational issues encountered. In general, the target for this SHADOZ station is to launch at least two ozonesondes per month, which makes a total of 24 launches per year.

**Figure 1.** Number of ozonesondes launched at Irene from 2000 to 2015, expressed (**a**) monthly and (**b**) annually.

Figure 2. shows O3 vertical profiles for the troposphere (Figure 2a) and stratosphere (Figure 2b). The data was grouped into three categories namely 2000–2003, 2004–2007, 2012–2015, and the O3 composite was calculated at kilometre intervals from 1 km to 28 km. The data used in Figure 2 also include 12.4% of ozonesondes that didn't reach 28 km. Therefore, 87.6% of ozonesondes launched over the study period reached 28 km. Three vertical profiles (2000–2003, 2004–2007 and 2012–2015) were compared and the difference between 2000–2003 and 2004–2007 and between 2000–2003 and 2012–2015 was calculated to determine O3 variation over the years. The percentage decrease for 2004–2007 profiles were calculated from O3 difference between 2000–2003 and 2004–2007, similarly, the percentage decrease for 2012–2015 was calculated from 2000–2004 and 2012–2015 O3 difference. A continuous decrease from 2004–2007 and 2012–2015 was apparent at the height between 12 km and 28 km, with highest decrease occurring at a height between 14 km and 18 km. An average decrease of 6.0% and 9.1% was observed at a height between 12 km and 26 km for 2004–2007 and 2012–2015, respectively, when compared to the 2000–2003 period. These results are in agreement with a study by Ball et al. [10] which reported evidence from multiple satellite measurements that ozone in the lower stratosphere between 60◦ S and 60◦ N has indeed continued to decline since 1998.

**Figure 2.** Ozone troposphere (**a**) and stratosphere (**b**) profiles. The blue line indicates 2000–2003 averages, black line indicates 2004–2007 averages and red line indicates 2012–2015 averages. The error bars indicate the standard deviation.

Figure 3 shows the O3 seasonal vertical profiles averages for the troposphere and stratosphere. In the year 2012–2015, summer experienced high O3 increase from 5 km to 12 km (Figure 3a). On the other hand, 2000–2007 experienced high O3 mole fractions in the lower troposphere (Figure 3a,c,e,g), with the increase noted up to 8 km during the winter (Figure 3e). On the other hand, O3 decrease was observed in the upper troposphere and lower stratosphere, with 2012–2015 experiencing lower mole fractions compared to 2004–2007 and 2000–2003 (Figure 3b,d,f,h). In general, a continuous non–consistent O3 decrease occurred from the upper troposphere (above 16 km) to lower stratosphere (28 km) for all seasons. These results suggest that the observed O3 decrease in the lower stratosphere is independent of season. Whereas, O3 increase between 1.5 and 4 km could be related to increase in urban influence boundary layer precursors [46]. Such O3 precursors could originate from domestic heating and power stations. Maximum standard deviation is observed at altitudes closer to tropopause region. Such increase could be related to STE and other dynamic changes occurring in the tropopause region. This variation was observed to be lesser in summer when compared to other seasons. Sivakumar and Ogunniyi [38] reported similar observations of higher standard deviation closer to the tropopause height.

Tables 1 and 2 summarizes statistic of high O3 events that were obtained by using the 95th and 90th percentile composites and potential vorticity of 2 PVU as thresholds. All episodes that exceed the monthly 95th percentile composite and 2 PVU were automatically classified as high ozone events (Table 1). However, it was noted that more episodes could be identified when 90th percentile composite is used instead of 95th percentile (Table 2). The monthly average composites, 90th percentile and 95th percentile composites were calculated by using 2000–2015 Irene ozonesondes data. The maximum of the peak was defined as the highest ozone observed from a particular ozone profile at a particular altitude. In this case, it is the maximum of the profile dated on the first column. Delta ozone was defined as the difference between the maximum of a particular event observed at a particular altitude and 90th or 95th percentile composite profile. The altitude of the event was defined as an altitude where maximum values of O3 occurred. Based on observations indicated in Table 2, it is noted that stratospheric intrusions can reach 7 km altitude over Irene. Similar results of the occurrence of deep intrusions at 7 km altitude were reported at Reunion Island [47]. Clain et al. [47] used different PV thresholds (1.0, 1.5 and 2.0 PVU) on the study of STEs in Reunion Island. Their findings revealed that the number of detected STEs depends on the PV value and duration of back trajectories. About 9.9% STEs were detected using 2 PVU and 2 days back trajectories relative to 28.5% STEs detected using 1 PVU and 10 days back trajectories. In this study, 6.8% STEs were detected using 2 PVU and 90th percentile composite as a threshold. While 4.6% STEs were detected using 2 PVU and 95th percentile composite as a threshold. Figures 4–10b show more events of high PV propagation from the stratosphere to the troposphere. However, in most cases there were some data gaps of ozonesondes.

**Figure 3.** Ozone seasonal profiles for the troposphere (**a**,**c**,**e**,**g**) and stratosphere (**b**,**d**,**f**,**h**). Blue line indicates 2000–2003 averages, black line indicates 2004–2007 averages and red line indicates 2012–2015 averages. The error bars indicate the standard deviation.

**Table 1.** Summary of high ozone (ppb) events statistics using 95th percentile and potential vorticity (2 PVU) as thresholds.



**Table 2.** Summary of high ozone (ppb) events statistics using 90th percentile and potential vorticity (2 PVU) as thresholds.

**Figure 4.** (**a**) Ozone troposphere profile over Irene on 07 July 2004. (**b**) Potential vorticity at Isobaric surface over Irene on July 2004 (from MERRA-2). Solid line in Figure 4a indicates monthly O3 average composite, broken line indicates O3 event and dotted line indicate 90th percentile composite. Colours in Figure 4b indicates the level of potential vorticity, black indicates high potential vorticity of more than 3 Km−<sup>2</sup> kg<sup>−</sup>1s−<sup>1</sup> while yellow indicates potential vorticity of less than 2 Km−<sup>2</sup> kg<sup>−</sup>1s−1. Blue arrow in Figure 4b indicates the event that is associated with O3 profile in Figure 4a.

**Figure 5.** (**a**) Ozone troposphere profile over Irene on 15 September 2004. (**b**) Potential vorticity at Isobaric surface over Irene on July 2004 (from MERRA-2). Solid line in Figure 5a indicates monthly O3 average composite, broken line indicates O3 event and dotted line indicate 90th percentile composite. Colours in Figure 5b indicates the level of potential vorticity, black indicates high potential vorticity of more than 3 Km−<sup>2</sup> kg<sup>−</sup>1s−<sup>1</sup> while yellow indicates potential vorticity of less than 2 Km−<sup>2</sup> kg<sup>−</sup>1s−1. Blue arrow in Figure 5b indicates the event that is associated with O3 profile in Figure 5a.

**Figure 6.** (**a**) Ozone troposphere profile over Irene on 26 August 2005. (**b**) Potential vorticity at Isobaric surface over Irene on July 2004 (from MERRA-2). Solid line in Figure 6a indicates monthly O3 average composite, broken line indicates O3 event and dotted line indicate 90th percentile composite. Colours in Figure 6b indicates the level of potential vorticity, black indicates high potential vorticity of more than 3 Km−<sup>2</sup> kg<sup>−</sup>1s−<sup>1</sup> while yellow indicates potential vorticity of less than 2 Km−<sup>2</sup> kg<sup>−</sup>1s−1. Blue arrow in Figure 6b indicates the event that is associated with O3 profile in Figure 6a.

**Figure 7.** (**a**) Ozone troposphere profile over Irene on 12 April 2006. (**b**) Potential vorticity at Isobaric surface over Irene on July 2004 (from MERRA-2). Solid line in Figure 7a indicates monthly O3 average composite, broken line indicates O3 event and dotted line indicate 90th percentile composite. Colours in Figure 7b indicates the level of potential vorticity, black indicates high potential vorticity of more than 3 Km−<sup>2</sup> kg<sup>−</sup>1s−<sup>1</sup> while yellow indicates potential vorticity of less than 2 Km−<sup>2</sup> kg<sup>−</sup>1s−1. Blue arrow in Figure 7b indicates the event that is associated with O3 profile in Figure 7a.

**Figure 8.** (**a**) Ozone troposphere profile over Irene on 31 July 2013. (**b**) Potential vorticity at Isobaric surface over Irene on July 2004 (from MERRA-2). Solid line in Figure 8a indicates monthly O3 average composite, broken line indicates O3 event and dotted line indicate 90th percentile composite. Colours in Figure 8b indicates the level of potential vorticity, black indicates high potential vorticity of more than 3 Km−<sup>2</sup> kg<sup>−</sup>1s−<sup>1</sup> while yellow indicates potential vorticity of less than 2 Km−<sup>2</sup> kg<sup>−</sup>1s−1. Blue arrow in Figure 8b indicates the event that is associated with O3 profile in Figure 8a.

**Figure 9.** (**a**) Ozone troposphere profile over Irene on 16 April 2014. (**b**) Potential vorticity at Isobaric surface over Irene on July 2004 (from MERRA-2). Solid line in Figure 9a indicates monthly O3 average composite, broken line indicates O3 event and dotted line indicate 90th percentile composite. Colours in Figure 9b indicates the level of potential vorticity, black indicates high potential vorticity of more than 3 Km−<sup>2</sup> kg<sup>−</sup>1s−<sup>1</sup> while yellow indicates potential vorticity of less than 2 Km−<sup>2</sup> kg<sup>−</sup>1s−1. Blue arrow in Figure 9b indicates the event that is associated with O3 profile in Figure 9a.

**Figure 10.** (**a**) Ozone troposphere profile over Irene on 25 November 2015. (**b**) Potential vorticity at Isobaric surface over Irene on July 2004 (from MERRA-2). Solid line in Figure 10a indicates monthly O3 average composite, broken line indicates O3 event and dotted line indicate 90th percentile composite. Colours in Figure 10b indicates the level of potential vorticity, black indicates high potential vorticity of more than 3 Km−<sup>2</sup> kg<sup>−</sup>1s−1while yellow indicates potential vorticity of less than 2 Km−<sup>2</sup> kg<sup>−</sup>1s−1. Blue arrow in Figure 10b indicates the event that is associated with O3 profile in Figure 10a.

#### *3.1. High O3 Events*

In this study, events with high O3 mole fractions exceeding monthly 90th percentile composites were selected and discussed in terms of PV retrieved using MERRA-2 reanalysis system. As indicated on a PV chart (PV plotted on an isotropic surface) in Figure 4, values outside the contour lines appear relatively low with PV values of about <sup>−</sup>1.9 <sup>×</sup> 10−<sup>12</sup> Km−<sup>2</sup> kg−1s−<sup>1</sup> compared with values inside the contour lines, which have PV values ranging between <sup>−</sup>4.7 <sup>×</sup> <sup>10</sup>−<sup>6</sup> to 1.9 <sup>×</sup> <sup>10</sup>−<sup>6</sup> Km−<sup>2</sup> kg-1s−1. Higher PV values in the troposphere indicate air mass of stratospheric origin due to increased static stability. PV is generally negative in the Southern Hemisphere (SH) and is usually multiplied by negative one (−1) to appear positive [48,49]. In general, higher values of PV are found in the stratosphere than in the troposphere.

Stratosphere polar vortex forms during autumn in Southern Hemisphere [50]. Moreover, with regards to the geographic position of Irene, this location experiences anticyclonic gyre due to midlatitude westerly waves that occur in autumn and winter [51]. These anticyclonic gyres are responsible for increase in pollutant concentrations for a long period [14]. Two possible stratosphere– troposphere O3 autumn events (12 April 2006 and 16 April 2014) were selected for discussion.

For the purpose of this study, three possible stratosphere–troposphere O3 winter events (7 July 2004, 26 August 2005 and 31 July 2013) were selected for discussion. Generally, one may conclude that activities such as anticyclone patterns, domestic usage of biofuels for heating, power generating plants and STEs are the cause of higher O3 enhancement during this period in Irene [51]. Furthermore, during this season, stratospheric polar vortex starts to be very active in polar region [50].

Also, in this study, two high O3 events (15 September 2004 and 25 November 2015) were selected in spring and discussed. It is well known that the spring O3 enhancement events can either be caused by biogenic emissions, biomass burning and lightning production or a combination of them all [11]. However, the occurrence of STEs is dominant in winter and spring over the study area [14], and during this time, the stratospheric polar vortex is more active. According to Clain et al. [47], it is possible that O3 mole fractions related to stratosphere intrusion can be influenced by climatological O3 background.

#### *3.2. Case Studies on High Ozone Events*

Figures 4a–10a show events that were selected for the case study. These events are part of the fifteen high O3 events that were identified over the study period. Since these events took place in the month of April, July, August and September, 90th percentile composites for these months were used as thresholds. On these days, high O3 peaks exceeding the monthly 90th percentile composites were observed between 9 km and 11 km. Figures 4b–10b show MERRA-2 potential vorticity plotted against the Isobaric surface between 70 hPa and 400 hPa. These vertical PV slices are for the whole month for the selected events, and averaged to the closest latitude and longitude to Irene SHADOZ site. As indicated on the PV slices, there are several episodes observed in these months where air masses with higher PV of approximately 3.0 PVU propagated from the lower stratosphere (70 hPa) to the upper troposphere (400 hPa). These events are shown as the downward tongues on the PV slices. However, for the purpose of this study, we focus on the time scale closer to the event dates. Blue arrows in PV slices indicate the events that are associated with high O3 in vertical profiles. As indicated by O3 vertical profiles, the observed high O3 events coincides with high PV observed in the higher troposphere. Therefore, it can be reasonably concluded that the observed high O3 in the upper troposphere could be of stratospheric origin.

#### *3.3. Dynamical Context Using MIMOSA Model*

There are several studies that have shown that the dynamics of the Southern Hemisphere polar vortex has an influence in the nearby surrounding structures (upper troposphere and stratosphere) of the Southern Hemisphere [28–30]. A useful method that can assist in profiling the isentropic transport across the dynamical barriers in the stratosphere is the MIMOSA (Modélisation Isentrope du transport Méso–échelle de l'Ozone Stratosphérique par Advection) model. MIMOSA model is a high-resolution advection contour model that is based on Ertel's potential vorticity which was developed at the Service d'Aeronomie by Hauchecorne et al. [52]. The advection is driven by ECMWF meteorological analyses at a resolution of 0.5◦ × 0.5◦. In the case of the PV, its slow adiabatic evolution is taken into account by relaxing the model PV towards the PV calculated from the ECMWF fields with a relaxation time of 10 days. Using this procedure, it is possible to run the model continuously and follow the evolution of PV filaments during several months. This model system enables the investigation of the contribution of the horizontal transport mechanism in the vertical distribution of ozone over high latitudes, mid-latitudes and subtropics. The model gives as an output the advected potential vorticity (APV) with a resolution of 0.3◦ × 0.3◦ which is measured in potential vorticity units (PVU) which corresponds to 1 <sup>×</sup> <sup>10</sup>−<sup>6</sup> Km−<sup>2</sup> kg<sup>−</sup>1s−1.

In their recent study, Orte et al. [30] successfully showed the influence of the polar vortex over Rio Gallegos, Argentina by using the APV calculated from the MIMOSA high-resolution advection model. Having adopted a similar approach in this study, the influence of the dynamics of the polar vortex over Irene during the days where the STE was observed is also investigated using the APV outputs from the MIMOSA model. The APV maps assimilated using MIMOSA model for the 350 K isentropic level plotted for the 07 July 2004 (a), 15 September 2004 (b), 26 August 2005 (c), 12 April 2006 (d), 30 July 2013 (e), 31 July 2013, 16 April 2014 (f), and 25 November 2015 (g) are shown in Figure 11. The location of Irene is indicated by a black dot in the maps. The slices of APV for 350 K isentropic level which is equivalent to 12–13 km were selected because this is found to be an appropriate pressure level to investigate an STE event.

**Figure 11.** Advected Potential Vorticity (APV) maps assimilated with the MIMOSA model for the 350 K isentropic level (in PVU) and for 07 July 2004 (**a**), 15 September 2004 (**b**), 26 August 2005 (**c**), 12 April 2006 (**d**), 30 July 2013 (**e**), 31 July 2013 (**f**), 16 April 2014 (**g**), and 25 November 2015 (**h**). The black dot represents the location of the Irene site.

In general, during all these days which experienced STE process there is obvious passing of APV values with an averaged value of 8 PVU over Irene, South Africa. This is confirmed by blue tongue at the 350 K isentropic level that reflect higher PV values passing over Irene during the days of the STE that were profiled in this study. It can thus be reasoned that this isentropic transport seems to be responsible for the observed reduction of the O3 mole fractions at the lower stratosphere over Irene which takes place at the same time with the enhancement of ozone in the upper troposphere. The possible dynamical event which is well simulated by the MIMOSA model during the STEs presented here could be that the high APV values are transported from the high latitudes towards the tropics bringing air masses that contain lower ozone concentrations. There is an upward propagation of the middle atmosphere planetary waves in the high and mid-latitudes regions which results to downward propagation around the lower latitudes of the lower stratosphere [28]. The O3 mole fractions in the lower stratosphere are transported to the upper troposphere, and hence the observed STEs. The influence of the high latitude stratospheric air masses over Irene were also reported on by Semane et al. [28] via their study of the dynamics of the middle atmosphere during the winter of year 2002. Besides, this was a special winter in the southern hemisphere because of the unprecedented year 2002 major stratospheric warming [53,54].

Diab et al. [14] reported a tropopause folding event that occurs during the winter and spring transition period in Irene. The tropopause folding is a good indicator of a STE physical process [14]. Thus, with an improvement of SHADOZ data collection at Irene site since then, it is always important to investigate such a physical process in this study. Figure 12 shows monthly averaged composite of O3 vertical profiles measured at Irene for the year period from 2000 to 2015. These profiles were plotted for the height region ranging from 1 km to 15 km for January (Jan) to December (Dec). There is a general significant intrusion of higher O3 mole fractions which are sourced from the stratosphere which is observed in late winter and spring months. In their study, Diab et al. associated this O3 injection to middle troposphere with westerly winds, which marks the end of maritime season [14]. Also, the subtropical jet was reported to play a role in permitting ozone-rich stratospheric air to penetrate into the troposphere [55]. While most of the free troposphere over Irene was characterised by O3 mole fraction of approximately 55–60 ppb, the late winter months experience an increase of O3 concentration to approximately 80 ppb just above the planetary boundary layer. It is also worth noting that the spring season is the period where there are activities such as anthropogenic pollutants sourced from the Congo region and Highveld region biomass burning, and natural activities such as lightning from rainy season and biogenic activities [56]. On the other hand, a similar observation to that which was reported by Diab et al. [14], the tropical tropopause layer (TTL) with O3 mole fraction ranging between 95 ppb and 100 ppb at a height above 14 km from January to February, while it noticeably declined throughout the year, and approached its minimum altitude of 11 km in October.

**Figure 12.** Contour plot of Irene O3 mole fraction (in ppb) for the period, 2000 to 2015.

#### *3.4. Ozone Decline in Lower Stratosphere*

The recovery of O3 in the upper stratosphere has been well discussed [9,57–61]. However, the investigation of O3 recovery at different altitude starting from the lower stratosphere upwards still needs attention. This also arise because some recent studies seem to have reported that there may be a continuous decline of O3 at the lower stratosphere [10,62]. A recent study by Sivakumar and Ogunniyi [38] divided ozonesondes data into two categories, namely tropospheric (0–15 km) and stratospheric region (15–30 km) and reported O3 maximum occurrence between 22–27 km. Thus, in this study, we also investigate the O3 decline in the lower stratosphere by using Irene ozonesondes data. Ozone decline is calculated by using medians, 5th and 95th percentiles. Moreover, a composite was calculated at different altitudes (e.g., 13–15 km, 16–18 km and 19–21 km) of the upper troposphere and stratosphere. Rate of change was then calculated by fitting a linear trend on the graphs.

#### 3.4.1. Annual Changes at Different Altitudes

Table 3 summarises the statistics calculated for the medians, 5th and 95th percentiles. Layers corresponding to the upper troposphere (7–9 km) show a positive change of 0.33 ± 0.57, 0.19 ± 0.56 and 0.38 ± 0.88ppb/year for the median, 5th and 95th percentiles respectively. Similarly, at 10–12 km height, a positive change of 0.29 ± 0.58 and 0.24 ± 1.25 ppb/year was observed for the median and 95th percentile respectively. In contrast, a negative change of 2.58 ± 3.90, −0.59 ± 3.17 and −9.63 ± 9.27 ppb/year was observed at 16–18 km for the median, 5th and 95th percentile respectively. Similarly, negative changes were observed at 19–21 and 22–24 km for median, 5th and 95th percentiles. In summary, the results presented here indicate that there are negative changes in the lower stratosphere, while the upper troposphere shows a positive change. Similar observations of O3 decline in the lower stratosphere were reported by Granados–Munoz and Leblanc [37] when studying tropospheric O3 seasonal and long-term variability at the JPL–Table Mountain. Furthermore, Ball et al. [10] suggested that lower stratosphere decline contributes to the observed total column O3 decline. Therefore, the results presented here are consistent with the previous observations reported in literature [10,37].


**Table 3.** Statistical analysis of ozone at different altitudes.

#### 3.4.2. Seasonal Changes at Different Altitudes

Figure 13 indicates O3 changes calculated using linear regression at different altitudes during summer, autumn, winter and spring season. And, Table 4(a,b,c,d) provide summery statistics (Median, 5th and 95th percentiles) of these changes. Standard deviation values close to zero indicate O3 observed over the years was close to the calculated mean. The observations in these tables can be summarised as follows:

7–9 km layer: there was a negative change identified in autumn (−0.11±0.53 ppb/year). While there was a positive change observed in summer (0.53 ± 0.40 ppb/year), winter (0.85 ± 0.72 ppb/year) and spring (0.04 ± 0.57 ppb/year) for the medians. Similarly, there was a positive change observed for the 5th percentiles in summer (0.60 ± 0.58 ppb/year), autumn (0.40 ± 0.34 ppb/year) and winter (0.02 ± 0.17 ppb/year). While spring (−0.02 ± 0.22 ppb/year) showed a negative change. There was a negative change observed in spring (−0.54 ± 1.00 ppb/year), while a positive change was observed in summer (0.23 ± 0.78 ppb/year), autumn (0.68 ± 0.79 ppb/year) and winter (1.14 ± 0.94 ppb/year) for the

95th percentiles. Therefore, it can be concluded that an overall positive change is dominant in this layer for most of the seasons.

**Figure 13.** Ozone time series (2000–2015) of the 5th (red), median (blue) and 95th (black) percentile ozone mole fractions at different altitude, (**a**) summer 13–15 km, (**b**) summer 16–18 km and (**c**) summer 19–21 km, (**d**) autumn 13–15 km, (**e**) autumn 16–18 km, (**f**) autumn 19–21 km, (**g**) winter 13–15 km, (**h**) winter 16–18 km, (**i**) winter 19–21 km, (**j**) spring 13–15 km, (**k**) spring 16–18 km and (**l**) spring 19–21 km. Dashed lines represent the linear fit for each time series.

10–12 kmlayer: therewas a negative change observedin autumn for themedians (−0.03 ± 0.81 ppb/year) and for the 5th percentiles in summer (−0.07 ± 0.66 ppb/year), autumn (−0.13 ± 0.61ppb/year) and winter (−1.00 ± 0.85 ppb/year). Similarly, there was a negative change observed in winter (−0.96 ± 2.03 ppb/year) and spring (−0.36 ± 1.19 ppb/year) for the 95th percentiles. Whilst on the other hand, a positive change was observed in summer (1.09 ± 0.72 ppb/year) and autumn (1.17 ± 1.04 ppb/year) for the 95th percentiles.

13–15 kmlayer: with the exception of winter (−0.18±1.34 ppb/year) and spring (−0.50 ± 0.88 ppb/year), a positive change was observed in summer (0.91 ± 1.26ppb/year) and autumn (0.62 ± 0.67 ppb/year) for the medians. Similarly, a positive change was also observed in summer (0.59 ± 1.47 ppb/year), autumn (1.02 ± 0.81 ppb/year), winter (0.03 ± 0.93 ppb/year) and spring (0.24 ± 1.31 ppb/year) for the 5th percentiles. Contrasting with the 10–12 km layer, the 95th percentiles yielded negative change in all of the seasons.


**Table 4.** Ozone statistical summary at different altitudes and seasons.

16–18 km layer: with the exception of autumn (0.93 ± 2.53 ppb/year), negative changes were observed in summer (−0.28 ± 1.84 ppb/year), winter (−3.85 ± 6.26 ppb/year) and spring (−7.11 ± 4.99 ppb/year) for the medians. Similarly, there was a negative change observed in autumn (−0.87 ± 1.69 ppb/year), winter (−0.88 ± 3.31 ppb/year) and spring (−1.15 ± 5.73 ppb/year) for the 5th percentiles except in summer (0.56 ± 1.96 ppb/year). The 95th percentiles again showed negative changes in all of the seasons. Standard deviations of more than 1.5 ppb were observed in all seasons.

19–21 km layer: with the exception of autumn (1.40 ± 8.90 ppb/year), negative changes were observed for the medians in summer (−1.28±8.96 ppb/year), winter (−13.30 ± 23.48 ppb/year) and spring (−14.61 ± 10.94ppb/year). In addition to this, negative changes were also observed in all seasons for the 5th and 95th percentiles. With the exception of 5th percentiles in autumn, standard deviations in excess of 8.0 ppb were calculated in this layer with more variation observed during winter and spring.

22–24 km layer: There was a negative change observed in all seasons for the medians, 5th and 95th percentiles. A standard deviation of more 14.0 ppb was calculated in this layer with more variation in winter and spring. The observed high standard deviations suggest greater significance of changes within this layer.

#### **4. Summary and Conclusions**

This study examined Irene O3 profile data from 2000–2015 in order to identify high O3 events and to study O3 decline at different altitudes of the stratosphere. Monthly 90th percentile composites were used as a threshold to identify high O3 events. Furthermore, PV charts at isobaric level were used to identify high PV air mass of stratospheric origin (more than 2 PVU). Based on the observations, high O3 events were found to occur in all seasons. However, they were most prevalent in winter and spring. The results showed that high O3 of stratospheric origin can propagate down to 7 km over Irene. However, very few events were found to reach this altitude. The majority of events occurred between 9 km and 10 km from the earth surface. Based on the results obtained from the PV charts, high PV values of approximately 3 PVU were observed over Irene.

Furthermore, O3 data was grouped into three categories: 2000–2003, 2004–20007 and 2012–2015 to investigate possible long-term changes using monthly 5th percentile composites, monthly 95th percentile composites and monthly median composites. Troposphere and stratosphere O3 vertical profiles were generated for the three datasets (2000–2003, 2004–20007 and 2012–2015). Based on the vertical profile graphs, it was noted that the maximum standard deviation occurred in altitudes closer to the tropopause (approximately 17 km). This could be related to STE and other dynamic changes occurring in the tropopause region.

The annual changes presented in Table 3 show an O3 decline at 13–15 km for the 95th percentiles (–2.38 ± 3.28 ppb/year) while median and 5th percentile O3 decline started at the 16–18 km layer. A maximum decline was observed at 22–24 km for the medians (−16.16 ± 21.83 ppb/year), 5th (−21.19 ± 27.39 ppb/year) and 95th percentile (−14.81 ± 21.82 ppb/year). High O3 decline was observed at 19–21 km and 22–24 km in all seasons. The 95th percentiles show O3 decline at 13–15 km. While O3 decline was observed in winter and spring at 10–12 km layer for medians and 95th percentiles. In conclusion, high O3 of stratospheric origin can occasionally reach down as low as 7 km above Irene. However, 68.8% of these events were observed within the 9 km to 10 km region.

PV charts proved a very useful tool and showed the propagation of stratospheric air masses to the troposphere as further evidence of stratosphere O3 intrusion for the selected high O3 episodes in this study. These observations seem to indicate that STE events which are observed in over Irene are strongly driven by the dynamics of the Southern Hemisphere polar vortex.

O3 decline was observed mainly in the lower stratosphere (16–28 km). However, it was more dominant in winter and spring, while few events were observed in summer and autumn. Contrary to this, O3 increase was observed in the lower troposphere. These observations of O3 increase in the lower troposphere are in line with literature reports and were associated with an increase pollution in the lower troposphere.

**Author Contributions:** Conceptualization, T.M. and N.M.; methodology, T.M.; validation, T.M.; formal analysis, T.M.; N.M. and N.B.; investigation, T.M. and N.M.; resources, T.M.; writing—original draft preparation, T.M.; writing—review and editing, T.M.; N.M.; V.S.; G.C.; C.L.; visualization, T.M.; N.M.; and V.S. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded jointly by the CNRS (Centre National de la Recherche Scientifique) and the NRF (National Research Foundation) in the framework of the LIA ARSAIO and by the South Africa/France PROTEA Program (project No 42470VA).

**Acknowledgments:** We thank South African Weather Service and ECMWF for access to the dataset. Our extended gratitude goes to Hassan Bencherif for his meaningful support. The ozonesonde data used here was obtained from the SHADOZ website https://tropo.gsfc.nasa.gov/shadoz/.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Ozone Trends in the United Kingdom over the Last 30 Years**

#### **Florencia M. R. Diaz 1,2, M. Anwar H. Khan <sup>1</sup> , Beth M. A. Shallcross 3, Esther D. G. Shallcross 3, Ulrich Vogt <sup>2</sup> and Dudley E. Shallcross 1,\***


Received: 22 April 2020; Accepted: 19 May 2020; Published: 21 May 2020

**Abstract:** Previous work regarding the behaviour of ozone surface concentrations over many years in the United Kingdom had predicted that the frequency and severity of ozone episodes would become less marked in the future as a response to environmental regulations. The aim of this study is to extend these studies and compare the results with their predictions. The ozone data of 13 rural and six urban sites in the UK collected from the Department for Environment, Food and Rural Affairs over a period from 1992 to mid-2019 were used to investigate this behaviour. The yearly ozone exceedances (the number of hours that the ozone concentration exceeded the 50 ppbv limit) in the United Kingdom were found to have decreased over the last 30 years regardless of the type of site (rural or urban), showing that the adopted emission controls have so far been successful in the abatement of pollutant emissions. In the past three decades, the highest numbers of exceedances were reached in May regardless of the type of site. Furthermore, these episodes have become less frequent and less severe in recent years. In fact, the number of hours of exceedance is lower than that in previous decades, and it is almost constant throughout the week.

**Keywords:** ozone exceedance; urban site; rural site; human health

#### **1. Introduction**

There is an increasing interest in the study of the behaviour of tropospheric ozone concentrations over the past 30 years due to its role as a greenhouse gas with an estimated globally averaged radiative forcing of 0.4 <sup>±</sup> 0.2 Wm<sup>−</sup>2, as a component of smog and as a primary tropospheric source of the hydroxyl radical (OH), which is a dominant tropospheric oxidant determining the lifetime of trace gases [1,2]. As a pollutant, ozone is corrosive and severely damaging to plants, trees and even buildings [3–5]. It has also been shown to have serious health effects on humans, particularly affecting the respiratory, cardiovascular and central nervous systems [6,7]. For example, it can cause irritation; it reduces the function of the lungs and promotes susceptibility to respiratory infections [8].

Tropospheric ozone is not directly emitted; it is formed as a secondary pollutant in the boundary layer by the reaction of primary pollutants (e.g., nitrogen dioxide and hydrocarbons) in the presence of sunlight. Because of this, its abatement depends on various parameters related to the emissions of these pollutants. Furthermore, ozone is transboundary [9], i.e., emissions from distant locations can contribute to its formation at a specific site; regulating the concentration of this pollutant requires international efforts. Ozone formation depends on the VOC–NOX ratio [10]. In urban areas, ozone concentrations are expected to rise due to its formation through NOX photochemistry, which originates

from road traffic exhaust, particularly primary diesel-fuelled vehicles at low speeds [11]. Ozone extremes have been found to have decreased during the last decade due to the substantial reduction of NOX emissions in response to protocols [12–14]. Rural background ozone concentrations have been found to be higher than those in urban backgrounds [15] due to two well established reasons, one being the (northern hemispheric) ozone baseline and the second being the presence of more NOX in urban sites actively scavenging ozone. A decrease in these pollutants would then lead urban sites to start behaving like their rural counterparts, and ozone concentrations would increase over the years as was found in [16].

Given that the formation of ozone depends on the concentration of the primary pollutants, a decrease in the concentration of pollutants like NOX causes ozone to decrease under the NOX-limited regime, while in cities (under the VOC-limited regime), a decrease in NOX leads to an increase in ozone levels. The Tropospheric Ozone Assessment Report (TOAR) showed that there is no clear global pattern for surface ozone changes since 2000, with increasing and decreasing trends in both polluted and remote sites [17]. Environmental policies have already been established to decrease the emissions of these pollutants. In Europe, for example, the Directive on Ambient Air Quality and Cleaner Air (2008/50/EC), adopted in 2008, aims to assess and manage the concentration thresholds of pollutants by setting limits [18]. If these limits are exceeded then authorities are required to implement plans to decrease these concentrations, as well as establish sanctions. These limits have also been observed by individual countries, and local objectives have been established. In the case of ozone, the objective limit in Europe is an 8 h mean of 60 ppbv (not to be exceeded more than 25 times a year, averaged over three years) [19], whereas in the United Kingdom, this limit is 50 ppbv (not to be exceeded more than 10 times a year) [20].

In order to understand what these policies have achieved in the reduction of the levels of the primary pollutants over the years, studies on ozone concentration trends in the United Kingdom have already been made [16,21–24]. This study is an update to these previous studies, addressing a more recent period and focusing on specific measuring sites throughout the UK. This study uses the freely available air quality data from the Department for Environment, Food and Rural Affairs (Defra) on selected rural and urban sites to address ozone exposure and understand its impact on human health. The aim of this study is to compare the results obtained in this work with the predictions made by the previous studies and determine if the implemented policies have succeeded or need to be amended. This study is thus a benchmark to indicate whether these policies have accomplished their objectives to date, or if pollution in the UK has worsened over the past three decades. Therefore, we analysed the ozone exceedances over the time period of 1992 to 2019 at 13 rural and six urban sites in the United Kingdom using the Defra ozone archive data. The decadal, yearly, monthly, daily and hourly variations of ozone exceedances are discussed in the study. We also analysed the magnitude and frequency of the most severe ozone episodes in rural and urban sites over the last 30 years.

#### **2. Methodology**

The source of the data for several measurement sites distributed throughout the UK was the Department for Environment, Food and Rural Affairs' (Defra's) Automatic Urban and Rural Network (AURN) data archive [25]. AURN is the UK's largest monitoring network comprising 150 sites and reporting hourly measurements of NOX, SO2, ozone, CO and particulate matter. We selected 13 rural and six urban sites in this study, which was based on the availability of the ozone data covering the last three decades (Figure 1). All these selected sites have an annual coverage of > 80% validated hourly data for ozone over the period from January 1992 to June 2019.

Ozone and NOX measurements were carried out by UV photometry and chemiluminescence, respectively, following the guidelines of the European Committee for Standardisation [26]. Data were validated on an ongoing basis by manual review to exclude any errors due to instrument malfunctions or faulty calibrations [27]. The ozone and NOX data (in μg/m<sup>3</sup> at 20 ◦C and 101.3 kPa) for each site were archived at the National Air Quality Information Archive [25]. The uncertainty (expressed as a

95% confidence level) of the measurement datasets and sites was around 15% [28]. The analysis was centred around finding the number of hours the ozone concentration exceeded the 50 ppbv limit (also known as an exceedance) in every selected site for a period from January 1992 to June 2019. These dates were chosen based on the date when urban sites began reporting ozone concentrations (rural reports go back as far as 1986) and the date when this study was first started. A linear regression method was used to estimate the trends in magnitude. Statistical significance is based on a *p* < 0.001 and the trends are reported with 95% confidence internals. Data are then analysed by type of site in order to find the frequency and magnitude of ozone episodes, as well as how trends have changed over the past three decades. Furthermore, in order to establish the causes for these pollution episodes, a meteorological back-trajectory model derived from the NOAA on-line trajectory service was used [29].


**Figure 1.** Distribution of the measurement sites (both rural and urban) in the UK.

#### **3. Results and Discussion**

#### *3.1. Three-Decadal Trend of Ozone Mixing Ratios*

The yearly averaged maximum ozone mixing ratios for 13 rural sites and six urban sites were found to have decreased gradually over the last 30 years (Figure 2) at rates of 1.0 ppbv/y (1.2%/y, *p* < 0.001) and 0.68 ppbv/y (0.9%/y, *p* = 0.002), respectively (the individual sites' maximum ozone and average ozone trends can be found in Table S1). The year-to-year ozone exceedance variability is highly dependent on meteorology, which makes it hard to separate the trends caused by any other effects (e.g., reduced precursor emissions) [30,31]. However, an increasing trend in yearly average ozone was found at a rate of 0.13 ppb/y (0.5%/y, *p* < 0.001) for rural sites and 0.20 ppb/y (1.1%/y, *p* < 0.001) for urban sites (Figure 2). The increases in yearly average ozone can be explained by a decreasing trend in average NOX mixing ratios at a rate of 0.22 ppb/y (3.7%/y, *p* < 0.001) for rural sites

and 1.2 ppb/y (3.3%/y, *p* < 0.001) for urban sites, resulting in less NOX scavenging [14,32,33], but the magnitude of the increasing trend at rural sites is smaller than those obtained at urban sites, due to the strong dependency of the concentrations of ozone on the northern hemispheric ozone baseline in rural areas [16]. Similar results for the decrease in maximum ozone concentration and increase in background ozone concentration are found in European sites for the period of 1995 to 2014 [34].

**Figure 2.** The mixing ratios of yearly average ozone, average NOX and maximum ozone for (**a**) rural and (**b**) urban sites over the last 30 years. Note: The data points indicate the averages of all rural sites (**a**) and urban sites (**b**) of the yearly averaged data for the indicated time period. No available data for rural NOX from 1992 to 2002. Trends are based on linear regression fitting with 95% confidence intervals and *p* values.

#### *3.2. Yearly Variation of Ozone Exceedances over the Three Decades*

The yearly total ozone exceedances for all rural sites show an overall decrease in the past three decades with the exceptions of the years 1992, 1995, 1999, 2003, 2006, 2008 and 2018 (Figure 3a). The weather conditions in the UK in the years 2003 [35,36] and 2006 [37] were particularly extreme, leading to a dramatic increase in summer pollution exceedances. The same applies for 1992 [38], 1995 [39] and 2018 [40], although the conditions were not as extreme in these years. This can be observed in Figure 3 where a high number of exceedances was reported for all these years. However, the high exceedances for 1999 and 2008 were not characterized by extreme weather conditions. The high exceedances in 1999 are due to a particularly long episode at the end of July, explained by the persistent

anticyclonic conditions during this period preventing the bad air conditions from dissipating [23]. The high exceedances in 2008 were mainly a result of the influence of the site Strathvaich, which reached 1000 h of exceedance in 2008 (Figure 3a), a quarter of the overall hours of exceedance from every site for this year. Most of these exceedances were reported in the spring months (e.g., March, April and May) rather than summer months (e.g., June, July and August) when ozone episodes are expected to happen (see Figure S1b). There is a typical spring-time ozone maximum characteristic of the northern hemispheric baseline air [41]. This cycle, shown to achieve a maximum between March and April, is also found in Strathvaich, which is due to an unclassical behavior of the baseline cycle. This means that the high number of exceedances in 2008 is due to this baseline cycle rather than any influence from ozone precursor emissions, and thus, the maximum achieved in 2008 in rural sites is not due to any severe pollution episodes.

**Figure 3.** Yearly total exceedances of ozone in (**a**) rural and (**b**) urban sites over the last 30 years. Note that the total exceedance is the total number of hours the ozone concentration exceeded the 50 ppbv limit for (**a**) rural sites and (**b**) urban sites.

Similarly, for urban sites, the years with the highest exceedances were found to be 1995, 1999, 2003, 2006, 2008 and 2018 (see Figure 3b). In 1992, "peak years" were identified at rural sites, which had very few hours of exceedance at urban sites compared with the other years. The 2008 maximum in urban sites was also influenced by only one site, Leeds Centre, with over 350 h of exceedance (Figure 3b), more than half of the overall hours from each site for this particular year. In 2008, most of the total hours of exceedance in Leeds Centre were reported in May (see Figure S2b). During this month, two episodes took place, one lasting for 14 days, starting on the 5th and ending on the 18th, and another lasting for 8 days from the 20th to the 27th. The highest ozone mixing ratio was recorded on the 11th, reaching 86.6 ppbv. Both episodes were only seen at Cardiff Centre, but their durations were shorter, one lasting for 6 days and the other lasting for 3 days. Birmingham Centre also reported one of these episodes, one starting on the 5th and lasting for 6 days. The rest of the sites did not report either of the episodes; in fact, Belfast Centre, London Bloomsbury and Southampton Centre showed no consecutive exceedances in May. The high number of exceedances during this year is attributed to the episode that took place earlier in the month. A backward trajectory analysis was performed for the Birmingham Centre, Cardiff Centre and Leeds Centre sites in order to deduce the cause of this episode. The trajectories (see Figure S3), following a 96 h span on the day the highest mixing ratios were reached, pass through continental Europe prior to their arrival in the UK at 4 p.m. local time. Since the highest mixing ratio was recorded at Cardiff Centre, it can be assumed that this magnitude is due to the air parcel passing directly over London and transporting with it a high concentration of ozone precursors. It is likely that during this episode, there was an anticyclonic weather event and easterly flows, which transport pollution from continental Europe to the UK, favouring considerable ozone production [42].

The number of exceedances for the urban sites is much lower than that for the rural sites (Table 1) because in urban sites, NOX emissions are more prevalent, which cause ozone scavenging, resulting in lower ozone mixing ratios. From the 1990s to the 2000s, the ozone exceedances increased. A decrease in NOX from the 1990s to the 2000s (Figure 2) is likely to have led to less NO scavenging, and ozone is then supposed to have accumulated, increasing the number of hours of exceedance. However, from the 2000s

to the 2010s, the ozone exceedances decreased, possibly because of the reduction in ozone precursor emissions (e.g., anthropogenic VOCs) caused by the 1999 Gothenburg protocol, which reduced VOC emissions in the UK by ~40% from the 2000s to the 2010s [43]. Overall, the exceedances for both urban and rural sites were decreased by 27% and 30%, respectively, from the 1990s to the 2010s.

**Table 1.** The average exceedances over the last three decades.


Note: Average values accounting for the number of sites of each type.

#### *3.3. Seasonal Trend Variation of Ozone Exceedances over the Three Decades*

The rural and urban sites follow a seasonal exceedances trend, with a summer high and winter low throughout the three decades (Figure 4). The ozone episodes depend on the weather conditions and, above all else, sunlight, since its formation is led by photolysis. It is then logical to assume that in a period in which sunlight is constant and its duration is long, a larger amount of ozone would be formed and accumulated, increasing its concentration. Furthermore, ozone episodes are linked with summer anticyclonic weather, leading to warm and sunny conditions with low wind speeds, lowering the dissipation rate of pollutants. However, this trend is not the same for each decade. The seasonal exceedances have all tended to peak in May, but the way in which the trend emerges is different throughout the years. For example, in the 1990s and 2000s, the exceedances dropped gradually after reaching the May maximum, but in the 2010s, the exceedances dropped to a minimum in June, increased in July and dropped again in August (Figure 4). These discrepancies can be explained by the variable weather conditions from year to year. For example, the June minimum in the 2010s decade is solely explained by the different weather conditions; among the summer months, June is characterized by lower pressure and temperature, and higher rainfall in the UK [44]. From 2011 to 2016, the weather was consistent, with frequent rain and low temperatures [45–50]. However, in 2017, a hot spell was recorded in mid-June, with temperatures reaching up to 28 ◦C towards the south east of England [51]. In 2018, a particularly heavy rainfall event was recorded resulting from Storm Hector [52], and in 2019, rainfall, around 2.5 times heavier than average, was recorded towards the East Midlands [53]. It is well established that rainfall and elevated wind speeds have a cleaning effect on pollution emissions [54], and thus, heavy rainfall would result in fewer ozone exceedances.

**Figure 4.** Seasonal total ozone exceedances in (**a**) rural and (**b**) urban sites averaged for the decades of the 1990s, 2000s and 2010s. Total exceedance is calculated as the total number of hours at an ozone concentration ≥ 50 ppbv for each month in a year, yearly averaged for each decade.

#### *3.4. Daily Variations of Ozone Exceedances over the Three Decades*

The diurnal plots for rural and urban sites (Figure 5) show that the highest ozone exceedances are found to have been reached at around 4 p.m. local time (mid-afternoon) throughout the three decades, and the lowest ozone exceedances were reported in the early mornings at around 7 a.m. to 8 a.m.; the cycle is very consistent with that in the similar study of Garland and Derwent [55]. This daily maximum is related to the influence of photochemical reactions with air pollutants, whereas the daily minimum is related to both ozone sinking by reaction with NO2 [56] and ground deposition [57]. The night decadal exceedances up to 80 h for rural sites and up to 8 h for urban sites (Figure 5) can be explained by meteorological factors, i.e., the formation of an inversion layer at night trapping the ozone formed during the day or the wind speed and direction either facilitating the transport of ozone precursors or dispersing NO2 and impeding its reaction with ozone. The decreased night-time exceedances for urban sites compared with rural sites can be explained by the sink reaction with NO2, since this pollutant is more readily available in an urban background compared with in a rural background.

**Figure 5.** Hourly total ozone exceedances in (**a**) rural and (**b**) urban sites averaged for the decades of the 1990s, 2000s and 2010s. Total exceedance is calculated as the total number of hours at an ozone concentration ≥ 50 ppbv for each hour in a day, yearly averaged for each decade.

#### *3.5. Weekly Variations of Ozone Exceedances over the Three Decades*

Ozone exceedances for urban and rural sites follow a weekly pattern (Figure 6) since ozone formation depends on the temporal variations in precursor emissions. The trend is much the same as in previous studies [15,23], with the highest exceedances on weekends (e.g., on Saturday and Sunday) throughout the three decades. For the rural sites, there was a less marked variation in the exceedances, which remained constant throughout the whole week. However, for urban sites, the weekly variation is significant because of the strong variability in the emissions of precursors between the weekend and weekdays. The lower weekend NOX concentrations due to lower traffic emissions reduce ozone scavenging, resulting in significantly higher weekend ozone concentrations than those on weekdays for urban sites [58].

**Figure 6.** Daily total ozone exceedances in (**a**) rural and (**b**) urban sites averaged for the decades of the 1990s, 2000s and 2010s. Total exceedance is calculated as the total number of hours at an ozone concentration ≥ 50 ppbv for each day in a week, yearly averaged for each decade.

#### *3.6. Frequency and Magnitude of Ozone Episodes*

In order to determine the magnitude and frequency of the most severe ozone episodes in rural and urban sites, data were analysed for the years 1992, 1995, 1999, 2003, 2006, 2008, 2018 and 2019, years in which the numbers of exceedances were extremely high compared with those in other years. Exceedances were then considered for each summer month to determine the number of consecutive days ozone concentrations reached the limit of 50 ppbv.

The most severe ozone episodes in the rural and urban sites data (Table 2) show that the ozone concentrations with the greatest exceedances in May decreased over the years. For the urban sites, the number of consecutive days when exceedances were achieved were much lower compared with those for the rural sites. It is difficult to determine a single month and duration of episodes in this type of site due to the variable number of exceedances from one site to the other. After comparing the days in which each site reached a maximum ozone concentration for the longest period, it is deduced that three severe ozone episodes occurred in 2003, 2006 and 2019. This is because the highest concentrations reached in these years were far greater than those of previous years in every single site regardless of type. These events, defined as Case Studies 1–3, were evaluated for each site to assess their origin and magnitude (Table 3).


**Table 2.** Duration and magnitude of most severe episodes in rural and urban sites.

Note: <sup>a</sup> Two episodes in June lasting for 9 days each; <sup>b</sup> Two episodes in July lasting for 4 days and 7 days, respectively; <sup>c</sup> In some sites, a single long episode lasting for 26 days and in all other sites, two episodes lasting for 8 days and 9 days, respectively; <sup>d</sup> Two episodes in August lasting for 3 days and 4 days, respectively; <sup>e</sup> In some sites, a single long episode lasting for 27 days and in all other sites, two episodes lasting for 9 days and 10 days, respectively.

#### Case Study 1

The earliest weeks of August 2003 were particularly problematic in several European countries. A heat wave caused temperature records to be broken in France, Germany, Italy, Spain, the Netherlands and the UK. During August 9th and 10th, both England and Scotland broke their previous temperature records: 38.3 ◦C in England and 32.9 ◦C in Scotland [36]. This hot spell was characterized by anticyclonic conditions, where a high-pressure system moving around western Europe brought a hot, dry tropical continental air mass to the UK. This pattern occurred for much of the rest of the month [59]. Moderate to high concentrations of ozone were reported in most of the sites evaluated in this study. The highest concentrations were reported in Harwell (Ha), Lullington Heath (LH) and Southampton Centre (SC), with mixing ratios of over 103 ppbv reached in the afternoon of August 9th on each site. The period for this episode spans 8 consecutive days on average, starting between the 3rd and 4th of August.


**Table 3.** Magnitude and duration of case studies.

Note: N/A—no available data.

#### Case Study 2

Another record-breaking heat wave struck the UK during the summer of 2006. It was found that July 2006 was the warmest month on record in the UK [60], which was characterized by warm, sunny days associated with high pressure systems over northern Europe. This month was at least 1 degree Celsius hotter than in the previous case study, and during the first four days of the month, temperatures exceeded 30 ◦C across England and Wales, with the next few days back to more normal conditions. On July 11th, temperatures increased again, and an anticyclonic weather system became established over the UK. This middle part of the month was at its warmest and sunniest [60]. During this period, the ozone mixing ratios were even higher than in the previous case study, with the highest mixing ratios recorded in Lullington Heath (LH) at 114.2 ppbv and very high concentrations recorded in Aston Hill (AH), Harwell (Ha) and Cardiff Centre (CaC), all above 100 ppbv. The analysis of Case Studies 1 and 2 suggests that ozone formation and accumulation are directly dependent on the weather conditions.

#### Case Study 3

This episode was not characterized by a heat wave, in contrast to the previous two case studies. Even though it was described as the hottest Easter Weekend in the UK according to the BBC [61], the temperatures hardly reached 25 ◦C all over the Kingdom, and by the 26th of April, storm Hannah had already reached the UK and dissipated the bad weather conditions with winds of over 70 mph [62]. Moderate mixing ratios above 70 ppbv were recorded in all sites except for SC. The highest mixing ratio achieved was 86.3 ppbv in High Muffles (HM). It is evident that this case study also had the

lowest recorded mixing ratios out of the three. However, the duration of this episode was much longer, lasting an average of 13 consecutive days on rural sites and 5 days on the urban ones.

We used the back-trajectories to determine from where the ozone precursors were coming and how they were affecting the sites in Case Study 3. The trajectories were followed by an incoming air parcel on a 96 h span towards three selected sites—Eskdatemuir (Ek), Lough Navar (LN) and Aston Hill (AH)—from April 17th to April 24th, arriving at 4 p.m. (local time), when the highest ozone mixing ratios were recorded for these sites. The sites chosen for this analysis were selected due to their location; an air parcel arriving in LN and AH (western sites) must travel through the south of the UK and possibly pass through other sites in this area. The same applies to Ek, since it is in the north of England. The trajectories for Ek and AH seem to have the same behaviour throughout the entire duration of the episode, passing through east Europe, the north of Germany and finally arriving in England, but on the 24th, the trajectory arriving in AH shifts suddenly from Europe to the Atlantic Ocean (see Figure S4). On the other hand, the trajectories arriving in LN change from day to day, coming in from France at the beginning of the episode, looping to the Atlantic Ocean on the 21st and then shifting again and arriving from central Europe on the day the highest ozone mixing ratios were reached and for the rest of the episode (see Figure S4). The highest ozone mixing ratio was recorded in HM on the 22nd at 8 p.m. local time. This site is on the Ek trajectory every single day of the episode. On the days prior to the 22nd, the trajectory loops over east and central Europe before arriving in Ek from a south-easterly direction (Figure S4). On the 22nd, the trajectories from both AH and Ek loop on each other from east Europe and separate over north Germany before arrival in the UK (see Figure S4). This means that the incoming air parcel on this day is the same for both sites, and when it finally arrives in HM, it has been carrying ozone precursors from four different countries. Two days later, the Ek trajectory changed its origin from east Europe to central Europe (Figure S4). It is then evident that the reason why this episode had such an effect on rural sites is purely the origin and transportation of ozone precursor emissions and not exclusively the weather conditions.

#### **4. Conclusions**

In this study, data from the Department of Environment, Food and Rural Affairs (Defra) was analysed in order to discuss the behaviour of ozone concentrations in the UK over a period from 1992 to mid-2019. This was done to explore how the frequency and magnitude of ozone episodes has changed over the years as a response to environmental policies to reduce pollutant emissions. It was found that maximum annual-mean ozone exceedances have decreased over the last three decades regardless of the type of site. The ozone exceedances in rural sites were found to be higher than those in the urban sites due to the lack of NOX emissions to scavenge ozone. Ozone episodes have been shown to take place exclusively in the late spring and summer, when the sunlight is constant and the weather conditions prohibit pollution from dissipating. The diurnal variation of ozone exceedances shows a maximum during the mid-afternoon for both rural and urban sites. Additionally, ozone episodes usually are more frequent on the weekends due to a reduced removal effect of NOX (emissions are lower on the weekends). It was also found that the day (in which ozone episodes are more frequent) has shifted from decade to decade due to the decrease in NOX levels. In recent years, these episodes have become less marked, with exceedances remaining constant throughout the whole week regardless of the type of site. The adopted emission controls in both the UK and Europe have so far been successful in decreasing pollutant emissions. Ozone trends during the 2010s decade have become less visible, but further analyses must be carried out periodically with different methodologies in order to adopt and implement the policies to control the trend, be able to achieve the ozone objective limit of 50 ppbv and grant a better quality of life for both people and ecosystems.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2073-4433/11/5/534/s1, Table S1: The annual average ozone maximum and the trend of maximum ozone and average ozone for the period of 1992–2019. Figure S1: Monthly ozone exceedances in rural sites for last three decades, (a) 1990s, (b) 2000s and (3) 2010s. Figure S2: Monthly ozone exceedances in urban sites for last three decades, (a) 1990s, (b) 2000s and (3) 2010s. Figure S3: Trajectories arriving in Birmingham Centre (Bi), Cardiff Centre (CaC) and Leeds Centre (LeC) at 4 pm (local time) on 11 May 2008. Figure S4: Trajectories arriving in Eskdatemuir (Ek), Aston Hill (AH) and Lough Navar (LN) at 4 pm (local time) for the period of 17 April to 24 April 2019.

**Author Contributions:** F.M.R.D. and M.A.H.K. analyzed the data and wrote the paper; U.V. and D.E.S. conceived and designed the project; B.M.A.S. and E.D.G.S. analyzed the data. All authors have read and agreed to the published version of the manuscript.

**Funding:** DES and MAHK thank Natural Environment Research Council (NERC), Bristol ChemLabS and Primary Science Teaching Trust under whose auspices various aspects of this work were funded.

**Acknowledgments:** We thank the Department for Environment, Food and Rural Affairs (Defra) for supporting UK monitoring network data and National Oceanic and Atmospheric Administration (NOAA) on-line trajectory service for providing meteorological back trajectories.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*
