Multi-year variation of near-surface ozone at Zhongshan Station, Antarctica

The online observation system of atmospheric composition were installed at the Swan Ridge (69°22′12″ S, 76°21′49″ E, 18.5 m a.s.l.), northwest of the Nella fjord. Strong E or ENE winds (45°–135°) dominate here, occupying 92% of time (Wang et al 2011). It implies human pollution from ZOS is weak.

In January 2008, an EC9810A ultraviolet photometric absorption O₃ analyzer was deployed to monitoring near-surface O₃, with an EC9811 as calibrator. Unfortunately, the instrument encountered a failure in January 2014, and the data was not credible till January 2016 when a TE49i ultraviolet photometric absorption O₃ analyzer and a TE49ips ultraviolet photometric absorption O₃ calibrator (Thermo Company) replaced. The measurement frequency was 1 min. Data are available at https://doi.org/10.11856/SNS.D.2021.001.v0. The meteorological monitoring including solar shortwave irradiance, wind speed was started in 1989 according to the World Meteorological Organization requirements. The Brewer monitoring including TCO was started in 1993 according to the World Ozone and Ultraviolet Radiation Data Centre.

In general, the zero-point, range and operating parameters of the O₃ analyzer should be checked regularly and before each operation. The calibration procedure followed the China’s environmental protection standard ‘ambient air—determination of O₃—ultraviolet method’ (HJ590-2010) (www.mee.gov.cn/gkml/sthjbgw/sthjbgg/201808/t20180815_451411.htm), which is stricter than the standards of the U.S. Environmental Protection Agency (www.epa.gov/ttn/amtic/files/ambient/pm25/qa/ QAHandbook-Vol-II.pdf.): the slope of the calibration curve ranged between 0.95 and 1.05, and the intercept ranged between −5 and 5 ppb. Every 3 months in ZOS, five standard concentrations of O₃ gas were generated for each calibration. From 2008 to 2020, 44 calibrations were made (e.g. excluding 2014 and 2015 data due to instrumental failure and it can be seen in supplement 1 available online at stacks.iop.org/ERL/17/044003/mmedia for the detailed results), and the correlation coefficient r was greater than 0.999 for the calibration quality of the observed data.

Furthermore, a variance test was applied to remove abnormal data based on the Laida criterion method, which assumes that the records obey a normal distribution. The formula is $\left| {xi-x} \right|\, > \,3\sigma$, where $xi$ is the measured value, $x$ is the time series mean and $\sigma$ is the standard deviation. After processing, 99.2% of the hourly mean data were retained from ZOS time series (excluding 2014 and 2015).

The HYSPLIT back trajectory analysis is a common application to determine the origin of air masses and establish source-receptor relationships (Stein et al 2015). It has been proven credible in previous Antarctica researches (e.g. Legrand et al 2009, Hara et al 2011). The TrajStat tools (MeteoInfoMap plugin for air mass trajectory statistic; Wang 2014) was be used to derive the model results in this study. Because O₃ usually has a half-life of several days in troposphere (Seinfeld et al 1998), the backward trajectory starting height was set at 100 m above the surface, and the total run time was 120 h for each trajectory. Each run was performed at time intervals of 24 h (00:00). In order to explore the major path ways of air mass arriving at the study area, the clustering analysis was carried out based on Euclidean distance. The final cluster number was 4, which determined by the point of inflexion on a curve of total spatial variance vs. number of clusters.

2.3.1. The PSCF model

The potential source contribution function (PSCF) model has often been applied to locate air masses associated with high levels of near-surface O₃ at different sites (Dimitriou and Kassomenos 2015, Sharma et al 2017, Yin et al 2017). In the summer over the Antarctic Plateau, higher PSCF values may point to sources of NO_X released from the snow that support O₃ production. (Cristofanelli et al 2018). The PSCF values for the grid cells in the study domain were calculated by counting the trajectory segment endpoints that terminated within each cell (Ashbaugh et al 1985). If the total number of end points that fall in a cell is ${n_{ij}}$ and there are ${m_{ij}}$ points for which the measured O₃ parameter exceeds a criterion value selected for this parameter, then the PSCF, can be determined as

$$\begin{equation}{\text{PSC}}{{\text{F}}_{ij}}\, = \,\frac{{{m_{ij}}}}{{{n_{ij}}}}.\end{equation} \tag{ 1 }$$

The cells with high PSCF values are associated with the arrival of air parcels at the receptor site, which has near-surface O₃ concentrations higher than the criterion value (i.e. the criterion threshold is the average concentration of non-OEEs (NOEEs) and OEEs). These cells are indicative of areas with ‘high potential’ contributions of the constituent. However, the identical PSCF ${n_{ij}}$ values can be obtained from cells with very different counts of back-trajectory points. To explain the uncertainty due to the low values of ${n_{ij}}$, the PSCF values were scaled by a weighting function ${W_{ij}}$ (Polissar et al 1999). The weighted PSCF (WPSCF) values when the total number of endpoints in a cell was less than approximately three times the average number of end points per cell. In this case, ${W_{ij}}$ was set as follows:

$$\begin{align}{W_{ij({\text{OEE}})}}{} = {} \left\{ {\begin{array}{*{20}{c}} {1.00} \\ {0.70} \\ {0.42} \\ {0.05} \end{array}\begin{array}{*{20}{c}} {} \\ {} \\ {} \\ {} \end{array}\begin{array}{*{20}{l}} {{n_{ij}} > 80{\text{Nave}}} \\ {80{\text{Nave}} > {n_{ij}} > 20{\text{Nave}}} \\ {20{\text{Nave}} > {n_{ij}} > 10{\text{Nave}}} \\ {{\text{Nave}} > {n_{ij}}} \end{array},} \right.\end{align} \tag{ 2 }$$

where ${\text{Nave}}$ represents the mean ${n_{ij}}$ of all grid cells. The WPSCF values were obtained by multiplying the original PSCF values by the weighting factor.

2.3.2. The CWT model

In the PSCF model, grid cells can have identical PSCF values when sample concentrations are slightly higher or much higher than the criterion value, making it difficult for the PSCF model to distinguish between strong pollution sources and weak pollution sources. In contrast, the concentration-weighted trajectory (CWT) model can well distinguish between strong and weak pollution sources. In the CWT model, each grid cell is assigned a weighted concentration by averaging the sample concentrations with associated trajectories crossing the grid cell, as shown below:

$$\begin{equation}{\text{WCW}}{{\text{T}}_{ij}}\, = \,\frac{{\sum\limits_{b = 1}^M {c_{\text{b}}}{\tau _{ijb}}}}{{\sum\limits_{b = 1}^{\text{M}} {\tau _{ijb}}}} \times {W_{ij}},\end{equation} \tag{ 3 }$$

where ${\text{CW}}{{\text{T}}_{ij}}$ is the average weighted concentration in the $ij$th cell, $b$ is the index of the trajectory, M is the total number of trajectories, ${c_b}$ is the concentration observed on arrival of trajectory $b$, and ${\tau _{ijb}}$ is the time spent in the $ij$th cell by trajectory l. A high value for ${\text{CW}}{{\text{T}}_{ij}}$ implies that air parcels traveling over the $ij$th cell would be, on average, associated with high concentrations at the receptor. Besides, the using of ${{\text{W}}_{ij}}$ in weighted concentration weighted trajectory (WCWT) same as that in WPSCF.

Figure 2 shows the time series of the monthly mean O₃ concentrations at ZOS from January 2008 to December 2020. O₃ was high in winter and low in summer. The lowest O₃ concentrations (12.2–19.9 ppb) were observed in January (2009, 2010, 2011, 2012, 2016, 2017, 2019, and 2020) and December (2008, 2013, and 2018). The highest O₃ concentrations (30.9–35.2 ppb) were observed in June (2013 and 2020), July (2008, 2009, 2010, 2011, 2012, 2016, and 2017), and August (2018 and 2019), which reflect the net O₃ accumulation effect during the Antarctic polar night. As the solar altitude rises and UV irradiation increases during early spring, near-surface O₃ starts to photolyze. Moreover, the NO_X , OH⁻ ion (hydroxide), and HO₂ concentrations remain extremely low due to the low solar radiation, and photolysis of O₃ still dominates, leading to a considerable decrease in O₃ concentration (Monks 2000).

Zoom In Zoom Out Reset image size

During the warm season (September–January), the O₃ concentration at ZOS exhibited large monthly fluctuations. This phenomenon has observed at many other coastal stations, such as the Syowa, Neumayer, Halley, and Arrival Heights stations (Ghude et al 2006, Helmig et al 2008, Wang et al 2011, Bian et al 2018). Surprisingly, abnormally high O₃ concentrations also occurred in September and October, sometimes exceeding the O₃ concentrations during the Antarctic polar night (as shown in figure 4). This phenomenon has been observed at other Antarctic inland stations, such as the SOP and Dome C (DMC) (Helmig et al 2008, Oltmans et al 2008).

To analyze the diurnal variability characteristics of O₃ at ZOS, the ΔO₃ (hourly growth rate of O₃) and diurnal variability of O₃ were calculated (figure 3(b)). The O₃ concentration at ZOS showed a clear diurnal cycle, with a maximum value at approximately 23:00 coordinated universal time (UTC) (4:00 local time (LT)) and a minimum value at approximately 9:00 UTC (14:00 LT). At ZOS, the monthly ΔO₃ varied little, ranging from −0.8 to 0.8 ppb h⁻¹, much smaller than the mean monthly ΔO₃ of 0.16 ppb h⁻¹ at DMC (Cristofanelli et al 2018). During the polar night, the daily mean wind speed remained stable, ranging from 6 to 8 m s⁻¹, but the ΔO₃ varied greatly at times (especially from 14:00 to 22:00 LT). This could only be attributed to the strong vertical exchange between the free atmosphere and PBL.

Zoom In Zoom Out Reset image size

Wind can promote mixing and transport between the PBL and the upper free atmosphere so that trace pollutants are continuously diffused in the horizontal and vertical directions. The diurnal variability of the wind speed may directly affect the reaction and transfer processes of near-surface O₃ (e.g. Argentini et al 2000). The O₃ concentrations continuously increase in spring and fall mornings with low wind speeds mainly because low wind speeds promote the accumulation of the O₃ precursors near the surface (Haman et al 2014). At the DMC, strong winds contribute to the dilution of snowpack surface-emitted O₃ precursors, which reduces the source contribution to the local photochemical process of O₃ formation (Legrand et al 2016). However, the wind speed at ZOS showed a notable diurnal cycle (figure 3(a)), which was positively correlated with the diurnal variability of O₃ (r² = 0.983, P< 0.01). This diurnal variability of O₃ at ZOS increased with wind speed, making it different from the diurnal variability of O₃ observed at DMC station. This difference may be attributed to the differences in geographical location and underlying surface between these two stations. Considering the diurnal variation characteristics of wind speed, the diurnal variability of O₃ at ZOS should be related to synoptic-scale air mass transport and the environmental background. In other words, the diurnal variability of O₃ at ZOS can represent the background characteristics of a large area.

Similar with the procedure in Cristofanelli et al (2018), the OEE day was selected based on a two-step method. First, the annual cycle of O₃ mean daily values was fitted with a sinusoidal curve. This represents an ‘undisturbed’ O₃ annual cycle, not affected by OEEs. In the second step, the probability density function (PDF) of the deviations from the sinusoidal fit was calculated, then a Gaussian fit was applied to the obtained PDF. It is proved that the deviations from the Gaussian distribution (calculated by using the Origin^® 9 statistical tool) can be used to identify observations affected by non-background variability (e.g. Giostra et al 2011). In order to obtain a threshold value for selecting non-background O₃ daily values possibly affected by ‘anomalous’ O₃ enhancements, we computed the further Gaussian fitting of PDF points falling above 1σ (standard deviation) of the Gaussian PDF. The intersection of the two fitting curves is taken as our screening threshold (table 1).

The OEE and NOEE days at ZOS were selected (figure 4). A total of 178 d over the 11 years period were affected by OEEs: 41 d in January (23.0%), 40 d in November (22.5%), 39 d in December (21.9%), 26 d in October (14.6%),and 11 d in September (6.2%). The frequency of OEE was less than 4% during other months (figure 5). Most of the OEEs at ZOS occurred in the warm season (September–January) and its intensity was much stronger than in the cold season. And we only found three OEEs during the polar night of 2018 (2%). Besides, we did not find obvious interannual variation characteristics of OEEs during 2008–2020.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Indeed, as suggested by Chevalier et al (2007), the comparison between hourly and daily O₃’s standard deviations were allowed to discriminate the relative contribution to the overall O₃ variability due to the diurnal variation process (PBL photochemistry and other processes) and the synoptic-scale variation process, i.e. weather change, synoptic-scale air mass transport, etc. The standard deviations based on hourly O₃ data during OEEs was minimal in January (2.9 ppb) and maximum in December (4.8 ppb). This indicated that the diurnal-scale process (i.e. PBL photochemistry and dynamics) played a role in enhancing O₃ variability during summer. However, even in December, the ratio between daily and hourly standard deviations suggested that 88.2% of O₃ variability at ZOS was associated with longer temporal scale processes.

Indeed, as suggested by Chevalier et al (2007), comparison of the hourly and daily O₃ standard deviations allows us to distinguish between the relative contributions of the diurnal variation processes (photochemical and other processes in the PBL) and the synoptic-scale variation processes (e.g. weather change, synoptic-scale air mass transport, etc) to the overall O₃ variability. Through calculation, we found that the annual mean contribution rate of the synoptic-scale variation processes at ZOS was 74%. In particular, the mean contribution rate of the synoptic-scale variation processes in the warm season (September–January) reached 81%. This indicates that the OEEs at ZOS are little affected by the diurnal variation processes and are mainly affected by the synoptic-scale variation processes.

In Antarctica, a decrease in TCO can enhance the incident UV radiation, leading to stronger photolysis of nitrates in snowpacks and thus the release of more NO_X . This could affect in situ photochemical production of O₃ in the long term (e.g. Jones and Wolff 2003). Figure 6 shows the monthly mean variability of OEE frequency, TCO, and solar shortwave irradiance at ZOS. The correlation coefficient (r) between the OEE frequency and solar shortwave irradiance reached 0.91 (P< 0.05). This indicates that as the solar shortwave irradiance increases during early summer, the photolysis of nitrates becomes stronger, thus increasing NO_X emission (Warneck and Wurzinger 1989, Honrath et al 2000). The corresponding chemical reaction process is as follows: NO₃ ⁻ + hv → NO₂ + O⁻, O⁻ + H → OH⁻. Then, NO₂ will produce O₃ through further photochemical reactions, resulting in favorable conditions for OEEs.

Zoom In Zoom Out Reset image size

As shown in figure 6, the correlation coefficient between the OEE frequency and TCO and the correlation coefficient between the OEE frequency and solar shortwave irradiance reached −0.93 and 0.99 (P< 0.05), respectively, during the period from July to October, providing strong evidence for the important role of stratospheric O₃ in modulating the near-surface atmospheric photochemical reaction processes by controlling UV radiation. The occurrence of an O₃ hole has gradually increased the incident UV radiation (the effect of the O₃ hole on solar shortwave irradiance can be determined by calculating the difference in the slope between the solar shortwave irradiance–time curve for the period from February to May (−0.25) and that for the period from July to October (0.29)), which is also beneficial to the formation of OEEs (Jones et al 2000, Jones and Wolff 2003).

To better study the effect of synoptic-scale processes in triggering OEEs, we calculated the total time that air masses spent over the plateau and coastal land areas (i.e. the area within the boundaries of 90°–66.5° S and 40°–160° E) before arrival at ZOS. The time spent by air masses over all land trajectories of OEEs was approximately 3.73% of the time spent by air masses over all trajectories and was approximately 78% of the time spent by air masses over all OEE trajectories (figure 7(a)). Clearly, the total time spent by air masses over the plateau and coastal land areas significantly impacted the OEEs at ZOS (figure 7(b)). The correlation between the two reached r = 0.82 and passed the 99% confidence level. The finding is in agreement with Legrand et al (2016) and Cristofanelli et al (2018). Just like the conclusion drawn in section 3.2, synoptic-scale air mass transport plays a pivotal role in triggering OEEs at ZOS.

Zoom In Zoom Out Reset image size

By clustering of OEE trajectories, we found that all air mass trajectories during OEEs originated from the coastal land areas of East Antarctica (the red, blue, yellow, and green lines in figure 8(a)). The O₃ concentrations of the four clusters were similar, all of them being between 20 and 31 ppb, with mean values from 25 to 27 ppb (figure 8(d)). In all months, the coastal cluster dominated during OEEs at ZOS (figure 8(c)). Furthermore, the pressures associated with the four clusters were higher than 550 hPa (figure 8(b)). These results indicate that most air masses were transported to ZOS through the troposphere over the Queen Mary Land, Wilkes Land, and Princess Elizabeth Land.

Zoom In Zoom Out Reset image size

The PSCF and CWT models were applied to find the potential pollution sources of O₃ during OEEs. The WPSCF values were high (>0.4) in the coastal area east of ZOS (from 76° to 150° E) (figure 9(a)). The WCWT values were greater than 20 ppb in the areas east of ZOS (figure 9(b)), which also indicated that the area east of ZOS was the O₃ source region that had the greatest impact on the O₃ at ZOS. The model results further confirmed that OEEs at ZOS were mainly caused by the transport of O₃-rich air masses from the atmosphere over the coastal land areas of the Princess Elizabeth Land, Wilkes Land, and Queen Mary Land.

Zoom In Zoom Out Reset image size

We applied the Stratosphere-to-Troposphere Exchange Flux (STEFLUX) tool to assess the possible contribution of STT to near-surface O₃ variability of ZOS, and to determine the measurement periods possibly affected by STT events (i.e. stratospheric air masses transferred down to the lower troposphere). We set the top lid of the target box at 300 hPa, and the following geographical boundaries: 68°–71° S and 75°–78° E. If at least 1 stratospheric trajectory crossed the 3D target box, then a ‘deep’ STT event at ZOS was detected (i.e. the STEFLUX ‘target box’; for further details on the methodology, see Putero et al 2016).

Figure 10 showed the annual (figure 10(a)) and monthly (figure 10(b)) distributions of STT events occurring at ZOS during the 2008–2018 (i.e. Since the STEFLUX tool are based on ERA-Interim data, the calculations were performed until 31 August 2018). The seasonal variation was not obvious, because of the low frequency of STT events. For example, the highest monthly frequency of STT events was only 0.9% (July). It was almost an order of magnitude lower than that of the OEEs. Therefore, it can be inferred that there was no direct link between the STT and OEE. This pattern was also reported in some previous studies (e.g. Stohl and Sodemann 2010, Cristofanelli et al 2018).

Zoom In Zoom Out Reset image size

However, some previous studies had considered the STT process is complex on Antarctica (Roscoe 2004, Mihalikova and Kirkwood 2013). The STEFLUX is only able to consider relatively ‘young’ events (i.e. 4 d old). This can lead to an underestimation of the real influence of STT at ZOS. In addition, STT events can play a role by transporting nitric acid, from stratosphere to the Antarctic atmosphere, thus indirectly affecting near-surface O₃ concentration and facilitating the occurrence of OEEs (Traversi et al 2014, 2017). For stations on the Antarctic coast, e.g. Neumayer, stratospheric air mass intrusions in latesummer/early autumn have been proven by long-termmeasurements of ¹⁰Be/⁷Be in surface aerosol particles (Elsässer et al 2011). At ZOS, through an observational analysis of the Δ¹⁴CO₂, it was also confirmed that the OEE of 6 January 2011 was affected by the STT (Zheng et al 2020). Therefore, it is important to carry out further studies to better assess this process.