The Impacts of Two Types of El Nio on Global Ozone Variations in the Last Three Decades

Stratospheric ozone concentrations directly affect levels of ultraviolet (UV) radiation reaching the Earth's surface and stratospheric temperature structure, while tropospheric ozone is considered to be a harmful air pollutant. Thus, the character, distribution and impact of ozone have been broadly studied (e.g., Chen et al., 1998, 2002, 2005, 2006; Hu and Tung, 2003; Zheng et al., 2003a, 2003b; Bian et al., 2005, 2006, 2007; Shi et al., 2007, 2010; Liu et al., 2009a, 2009b, 2011; Eyring et al., 2010). Meanwhile, the factors controlling ozone changes have always constituted an important research topic.

The canonical El Niño is the most significant mode of variation in the atmosphere. It can be derived from the first mode of an empirical orthogonal function (EOF) analysis of tropical Pacific Ocean SST (e.g., Rasmusson and Carpenter,1982; Trenberth, 1997; Zhang et al., 2009). Canonical El Niño can cause anomalous propagation and dissipation of ultralong Rossby waves to impact the stratosphere at middle latitudes (van Loon and Labitzke, 1987; Hamilton, 1993; Sassi et al., 2004; Garc\'ía-Herrera et al., 2006; Manzini et al., 2006; Taguchi and Hartmann, 2006; Camp and Tung, 2007; Garfinkel and Hartmann, 2007; Free and Seidel, 2009; Xie et al., 2011) and influences the troposphere by adjusting the spatial distribution of convection (Reid and Gage, 1985; Philander, 1990; Yulaeva et al., 1994; Newell et al., 1996; Chandra et al., 1998; Randel et al., 2000; Gettelman et al., 2001; Kiladis et al., 2001; Sassi et al., 2004). Thus, canonical El Niño can influence the global ozone distribution by causing these circulation changes.

Based on an analysis of a global chemistry-climate model simulation of the atmosphere forced by observed SST variations and Stratospheric Aerosol and Gas Experiment II (SAGE II) observations of atmospheric composition, (Randel et al., 2009) concluded that canonical El Niño events are tied to fluctuations in tropical upwelling. These fluctuations in tropical upwelling are linked to coherent anomalies in zonal mean temperature and ozone in the tropical lower stratosphere. Using a multiple regression method, (Hood et al., 2010) found that canonical El Niño is associated with reduced ozone in the lower stratosphere and increased ozone in the middle stratosphere. The response of stratospheric ozone to canonical El Niño has also been analyzed using a variety of chemistry-climate models (Fischer et al., 2008; Cagnazzo et al., 2009) and a chemical transport model (Sassi et al., 2004). Recently, (Randel and Thompson, 2011) identified canonical El Niño as one of the main factors controlling variations in stratospheric ozone. In the troposphere, canonical El Niño events are associated with dryness, downward motion and suppressed convection in the tropical western Pacific over Indonesia. These changes are associated with an increase in tropospheric column ozone both because dryness increases the chemical lifetime of ozone and because the suppression of convection leads to weakened vertical transport of near-surface ozone. By contrast, in the eastern Pacific, increased humidity and enhanced upward transport result in a negative tropospheric column ozone anomaly (e.g., Kley et al., 1996; Chandra et al., 1998; Hauglustaine et al., 1999; Thompson and Hudson, 1999; Fujiwara et al., 2000; Sudo and Takahashi, 2001; Thompson et al., 2001; Chandra et al., 2002, 2007, 2009; Ziemke and Chandra, 2003, Ziemke et al., 2010; Doherty et al., 2006; Oman et al., 2011). Moreover, several previous studies have pointed out that changes of stratosphere-troposphere exchange (STE) caused by canonical El Niño also significantly affect ozone concentrations in the extratropical troposphere (Langford et al., 1998; James et al., 2003; Zeng and Pyle, 2005).

The second leading mode of an EOF analysis of tropical Pacific Ocean SST variability from 1980 to 2010 shows two cold centers in the eastern and western Pacific and a warm center in the central Pacific (Fig. 1b), referred to as El Niño Modoki, compared with a cold center in the eastern Pacific and a warm center in the western Pacific in canonical El Niño events (Fig. 1a) (Ashok et al., 2007). Since the late 1970s, the number of El Nino Modoki events has been increasing (Ashok and Yamagata, 2009; Yeh et al., 2009). However, the impacts of El Niño Modoki on ozone changes do not receive a lot of attention.

Figure 1.  SSTAs in the tropical Pacific Ocean during (a) canonical El Niño events and (b) El Niño Modoki events. Solid lines represent positive values. Contour lines are ± 0.3^°C.

Canonical El Niño and El Niño Modoki events can lead to profoundly different effects on stratospheric temperature and circulation (Hu and Fu, 2009; Hu and Pan, 2009; Hurwitz et al., 2011a, 2011b; Lin et al., 2012; Xie et al., 2012; Zubiaurre and Calvo, 2012). This is because, in the mid-latitude stratosphere, the gradient patterns of sea surface temperature anomalies (SSTAs) in the two types of El Niño events can cause completely different patterns of propagation and dissipation of ultralong Rossby waves. Trenberth and Smith (2006, 2009) noted that the two types of El Niño result in distinct temperature anomalies in the high-latitude stratosphere in the Southern Hemisphere. (Xie et al., 2012) pointed out that zonal circulation changes are completely different in the two types of El Niño in the middle-high-latitude stratosphere, i.e., canonical El Niño events weaken and warm the north polar vortex but strengthen and cool the south polar vortex; however, El Niño Modoki events strengthen and cool the north polar vortex but weaken and warm the south polar vortex. (Ashok et al., 2007) found that a tripolar pattern of sea level pressure anomalies appears in the evolution of El Niño Modoki events where the centers of the tripolar pattern locate in the tropical western Pacific, central Pacific and eastern Pacific; and in El Niño Modoki events there are two anomalous Walker circulation cells (compared with the single cell related to canonical El Niño events).

Figure 2.  Longitude-latitude cross sections of zonally-averaged ozone (averaged over 2005-2010) from (a) ERA-Interim ozone and (b) Aura MLS ozone. Contour intervals are 1.8 ppmv. Total column ozone (TCO) distributions (c) based on ERA-Interim data averaged over the period 1980-2010 and (d) based on NIWA data averaged over the period 1980-2009. Contour intervals are 10 DU. (e) Time series of ERA-interim (solid line) and NIWA (dashed line) TCO averaged over (50^°S-50^°N, 180^°W-180^°E).

During El Niño Modoki events, the anomalies in the atmospheric circulation seem to affect the stratosphere and troposphere differently from those associated with canonical El Niño events. Thus, global stratospheric and tropospheric ozone may respond differently to the two types of El Niño events. This study not only investigates the different effects of the two types of El Niño on global ozone, but also tries to explain the mechanisms underlying these effects. The manuscript is organized as follows. Section 2 describes the data and numerical simulations used in the study. Section 3 discusses the relationships between patterns of stratospheric ozone changes and the two types of El Niño, and section 4 analyzes the variations in tropospheric ozone patterns associated with the two types of El Niño. Conclusions are presented in section 5.

Ozone is taken from monthly mean European Center for Medium Range Weather Forecasting (ECMWF) reanalysis data (ERA-Interim) between 1980 and 2010 (Simmons et al., 2007a, b; Uppala et al., 2008). ERA-Interim has assimilated satellite observations (reprocessed Global Ozone Monitoring Experiment data from the Rutherford Appleton Laboratory provides ozone profile information from 1995 onwards) at a 1.5^°× 1.5^° horizontal resolution and relatively high vertical resolution (37 levels) compared with the ERA40 reanalysis (23 levels) (Simmons et al., 2007a, b; Uppala et al., 2008). Figure 2 shows the longitude-latitude cross sections of zonally-averaged ozone (averaged over 2005-2010) from ERA-Interim and the Aura Microwave Limb Sounder (MLS), horizontal distributions of ERA-Interim total column ozone (TCO) (averaged over 1980-2010) and assimilated observed TCO from the National Institute of Water and Atmospheric Research (NIWA) (averaged over 1980-2009) (Bodeker et al., 2005), and the time series of ERA-Interim and NIWA TCO averaged for the latitude band of 50^°S-50^°N. The averaged latitude range of this limitation is dictated by the number of missing values in the NIWA TCO at high latitudes, so that ERA-Interim and NIWA TCO can only be compared at lower-middle latitudes. The spatial distribution and time series of ERA-Interim ozone agree well with ozone observed by the Aura MLS and NIWA. (Dragani, 2011) also compared ERA-Interim ozone with satellite data and concluded that the ERA-Interim data are reasonable and realistic. ERA-Interim ozone data are therefore suitable for diagnosing spatial and temporal ozone anomalies.

To identify occurrences of El Niño Modoki and canonical El Niño events, we use the monthly ENSO Modoki index, hereafter EMI, and the NINO3 index, hereafter N3I, respectively. The EMI is defined as follows, according to (Ashok et al., 2007):

where the subscripted brackets is the area mean SSTA over the central Pacific area [(SSTA) _C: (10^°S-10^°N, 165^°E-140^°W)], the eastern Pacific area [(SSTA) _E: (15^°S-5^°N, 110^°-70^°W)], and the western Pacific area [(SSTA) _W: (10^°S-20^°N, 125^°-145^°E)]. N3I is defined as the area-mean SSTA over the region (5^°S-5^°N, 150^°-90^°W), and is available at http://www.cpc.noaa.gov/data/indices/. Following previous studies (Weng et al., 2009; Zhang et al., 2010; Feng and Li, 2011), months with EMI values equal to or greater than +0.5^°C were identified as El Niño Modoki events. Similarly, months with N3I values equal to or greater than +0.5^°C were identified as El Niño events. Several strong and continuous El Niño samples have been selected for this study (Table 1).

The formulae to calculate the Brewer-Dobson (BD) circulation in a pressure coordinate system were given by (Edmon et al., 1980):

where θ is the potential temperature; a is the radius of the earth; is mean meridional wind; and is mean vertical velocity in pressure coordinates. Subscripts p and denote derivatives with pressure p and latitude , respectively. The overbar denotes the zonal mean and the accent denotes the deviations from the zonal mean value.

Figure 3 shows latitude-height cross sections of ozone anomalies above 100 hPa associated with canonical El Niño and El Niño Modoki events. Previous studies have shown that warm ENSO events weaken the northern polar vortex, which increases ozone concentrations at high latitudes in the Northern Hemisphere (e.g., Fischer et al., 2008; Cagnazzo et al., 2009), and strengthen BD-related dynamical transport from the upper troposphere to the lower stratosphere in the tropics (e.g., Sassi et al., 2004; Garc\'ía-Herrera et al., 2006; Manzini et al., 2006; Calvo et al., 2010), which reduces ozone concentrations in the tropical lower stratosphere (Randel et al., 2009; Hood et al., 2010). In agreement with these previous studies, canonical El Niño events lead to a positive ozone anomaly at high latitudes in the Northern Hemisphere and a negative ozone anomaly in the tropical lower stratosphere and a positive ozone anomaly at high latitudes in the Southern Hemisphere (Fig. 3a). The pattern of circulation anomalies associated with El Niño Modoki is totally different from those associated with canonical El Niño (Xie et al., 2012), and hence the pattern of ozone anomalies associated with El Niño Modoki is also totally different from those associated with canonical El Niño i.e. El Niño Modoki activities reduce ozone in the lower-middle stratosphere over almost the entire globe and in the upper stratosphere at high latitudes but increase ozone in the middle stratosphere at high latitudes (Fig. 3b). The ozone anomalies are arranged in a banded distribution.

Figure 3.  Latitude-height cross sections of zonally-averaged stratospheric ozone anomalies (%) composited for (a) canonical El Niño events and (b) El Niño Modoki events based on ERA-Interim data from 1980 to 2010. Contour intervals are 1% for ozone anomalies. Solid lines represent positive values; dashed lines represent negative values. Shaded areas are significant at the 95% confidence level (Student's t-test). The canonical El Niño anomalies are calculated using composites of the detrended and deseasonalized time series over the selected canonical El Niño events, just as the El Niño Modoki anomalies are calculated using composites over the selected El Niño Modoki events. For the selected canonical El Niño/El Niño Modoki events, please refer to Table 1. Before the composite analysis, the QBO signals in ozone were filtered out from time series using a 24-40-month band-pass filter.

(Randel and Wu, 1996) used a singular value decomposition (SVD) and regression analysis of observed stratospheric ozone from 1979 to 1995 to show that the first and second modes of the SVD are associated with the quasi-biennial oscillation (QBO) Together, these two dominant modes accounted for approximately 97% of the variance. More recently, (Randel and Thompson, 2011) reported that both QBO and ENSO have been major contributors to variations in stratospheric ozone. An EOF analysis of the stratospheric ozone 1980-2010 time series is used to isolate the QBO and ENSO signals in ozone over the past 30 years. The spatial and temporal variations of the first and second modes of this EOF analysis are shown in Fig. 4. The leading EOF spatial mode (EOF1; Fig. 4a), which accounts for 27% of the variance, corresponds to the El Niño Modoki pattern of stratospheric ozone anomalies (Fig. 3b). The variations of the leading principal component (PC1) are strongly correlated with EMI, with a linear correlation coefficient of 0.59 (Fig. 4b). The second spatial mode (EOF2; Fig. 4c) which accounts for 1% of the variance, is associated with the QBO pattern of ozone anomalies reported by previous studies (Randel and Wu, 1996; Lee and Smith, 2003; Crooks and Gray, 2005; Randel and Wu, 2007). (Randel and Wu, 1996) showed that the QBO signal in stratospheric ozone results from secondary circulations induced by the QBO. The variations of the second principal component (PC2) is significantly correlated with the QBO index (Fig. 4d), where the QBO index is defined as the normalized zonal wind at 30 hPa averaged between 10^°S and 10^°N. The linear correlation coefficient between PC2 and QBO index is 0.70 Figure 4 suggests that the impacts of El Niño Modoki on stratospheric ozone have exceeded those of the QBO over the past 30 years. The variations of mean ozone anomalies from 100 hPa to 1 hPa and the variations of PC1 are shown in Fig. 4e. The stratospheric ozone changes and PC1 are anti-correlated with a statistically significant linear correlation coefficient of -0.76 which is associated with El Niño Modoki activity. This strong anti-correlation implies that El Niño Modoki has played a key role in controlling variations of stratospheric ozone over the past three decades.

Figure 4.  Results of empirical orthogonal function (EOF) analysis of the 1980-2010 time series of zonally-averaged stratospheric ozone anomalies. The range for performing EOF analysis is from 100 hPa to 1 hPa and from 90^°S to 90^°N. The ozone anomalies were obtained by removing the annual cycle and linear trend from the original time series at each grid. The square-root of the cosine of latitude and square-root of the pressure interval was used for the weighting function in the EOF analysis. (a) The spatial pattern of the leading EOF mode (EOF1). Contour intervals are 0.01 ppmv. Solid lines represent positive values; dashed lines represent negative values. (b) The variations of the leading principal component (PC1; solid line) and EMI (dashed line). (c) As in (a), but for the spatial pattern of EOF2. (d) As in (b), but for the variations of PC2 (solid line) and the QBO index (dashed line). (e) The variations of PC1 (solid line) and mean stratospheric ozone anomalies (dashed line). All the time series are normalized.

Previous studies have shown that El Niño activities tend to strengthen the Brewer-Dobson (BD) circulation in the tropical lower stratosphere, which transports the tropospheric low-ozone air to the stratosphere and causes a local ozone decrease in the tropical lower stratosphere (Randel et al., 2009; Hood et al., 2010). The lead-lag correlations (LLCs) between the vertical component of BD circulation, diagnosed from National Centers for Environmental Prediction Reanalysis 2 (NCEP2) and ERA-interim data, and EMI and N3I index shown in Figs. 5a and b illustrate that the BD circulation variations are more closely linked with El Niño Modoki events than with canonical El Niño events during the past three decades. This may be the reason that El Niño Modoki can cause more significant ozone decreases in the middle-lower stratosphere compared with canonical El Niño. (Randel et al., 2009) and (Hood et al., 2010) noted that El Niño tends to reduce ozone in the tropical lower stratosphere and increase ozone in the tropical middle stratosphere. These studies did not differentiate between the two types of El Niño; however, it is evident that ozone anomalies in the tropical stratosphere are larger during El Niño Modoki periods than during canonical El Niño periods. These results suggest that El Niño Modoki events have a more significant impact on ozone distributions in the lower and middle stratosphere than canonical El Niño. The decreased lower stratospheric ozone and the increased middle stratospheric ozone observed during El Niño events may be mainly related to El Niño Modoki, but not canonical El Niño.

Figure 5.  (a) Lead-lag correlations (LLCs) between the vertical component of BD circulation and N3I (solid line) and EMI (dashed line) for the calculation of BD circulation from NCEP2 data. Mean area of BD circulation: (30^°S-30^°N, 180^°W-180^°E); 120 hPa-70 hPa. A positive lag means the EMI and N3I lead, and a negative lag means BD circulation leads. The dotted lines denote the 95% confidence level. (b) Same as (a), but the calculation of BD circulation is from ERA-interim data.

Figure 3 indicates that the effects of both types of El Niño on stratospheric ozone vary significantly in the spatial domain. Stratospheric column ozone anomalies are calculated to better estimate the net effects of El Niño events on total stratospheric ozone. These stratospheric column ozone anomalies are shown in Fig. 6. Stratospheric column ozone is calculated assuming that the stratosphere extends from the tropopause to 1 hPa. The tropopause is determined using the World Meteorological Organization (WMO) definition (thermal tropopause) in the tropics, the 2.5 PV pressure level (dynamic tropopause) at high latitudes, and a weighted average of the thermal and dynamic tropopauses (mixed tropopause) in middle latitudes. Canonical El Niño causes significant increases in stratospheric column ozone at middle-high latitudes in the Northern Hemisphere (Fig. 6a) because it weakens the Northern Hemisphere polar vortex. By contrast, canonical El Niño causes decreases in stratospheric column ozone in the tropics because it strengthens the BD circulation. Stratospheric column ozone also increases over the Southern Hemisphere middle-high latitudes, with the exception of a negative stratospheric column ozone anomaly between 60^°E and 180^°E. Stratospheric column ozone anomalies associated with El Niño Modoki are completely different from those associated with canonical El Niño. El Niño Modoki reduces global stratospheric column ozone (Fig. 6b). This is because El Niño Modoki events significantly decrease the lower-middle stratospheric ozone (Fig. 3b). The results also agree with the above if stratospheric column ozone is calculated assuming that the stratosphere extends from 200 to 1 hPa.

Figure 6.  Latitude-longitude cross sections of stratospheric column ozone anomalies (%) composited from 1980-2010 ERA-Interim data for (a) canonical El Niño events and (b) El Niño Modoki events. Contour intervals are 0.5% for ozone anomalies. Solid lines represent positive values; dashed lines represent negative values. Stratospheric column ozone is calculated assuming that the stratosphere extends from the tropopause to 1 hPa. Shaded areas are significant at the 95% confidence level (Student's t-test). Latitude-longitude cross sections of climatological lower-stratospheric meridional winds averaged from 100 hPa to 30 hPa based on (c) ERA-Interim data from 1980 to 2010. Contour intervals are 0.8 m s^-1. Solid lines represent northerly vectors; dashed lines represent southerly vectors. (d) Latitude-longitude cross section of anomalies of lower-stratospheric meridional wind composited for El Niño Modoki events based on ERA-Interim data from 1980 to 2010. Contour intervals are 0.2 m s^-1. Solid lines represent positive values; dashed lines represent negative values. (e) Same as (d), but for upper-tropospheric-averaged (300-100 hPa) geopotential. Contour intervals are 50 m² s^-2. (f) Same as (d), but for lower-stratospheric-averaged (100-50 hPa) vertical velocity. Contour intervals are 0.00005 Pa s^-1.

It is noteworthy that El Niño Modoki causes a more significant decrease in stratospheric column ozone over the tropical eastern Pacific (60^°-120^°W) than over other tropical areas (Fig. 6b). This decrease is particularly pronounced in the Southern Hemisphere tropical eastern Pacific. Further analysis of reanalysis data indicates that the decrease of lower-stratospheric ozone (averaged between 100 hPa and 30 hPa) during El Niño Modoki periods is most significant over the tropical eastern Pacific, particularly the Southern Hemisphere tropical eastern Pacific (not shown). The spatial pattern of anomalies in lower-stratospheric ozone concentrations is similar to that of anomalies in stratospheric column ozone.

Dynamical transport plays an important role in controlling the spatial distribution of ozone in the stratosphere. Figure 6c shows climatologies of lower-stratospheric meridional wind and El Niño Modoki anomalies of lower-stratospheric meridional wind (Fig. 6d) averaged between 100 hPa and 30 hPa from reanalysis data. It illustrates that the meridional wind is directed from middle-high latitudes to lower latitudes in both hemispheres throughout the longitude band of 60^°-120^°W (Fig. 6c). We found the equatorward wind in this longitude band is further enhanced during El Niño Modoki events, particularly in the Southern Hemisphere (Fig. 6d). (Renwick and Revell, 1999) pointed out that these significant meridional wind anomalies in middle-high latitudes are likely to be related to blocking over the south Pacific and Rossby wave propagation during ENSO events. It is indeed found that there is a positive anomaly of geopotential, which represents an abnormal high pressure center, near the longitude of 120^°W in the Southern Hemisphere high latitudes (Fig. 6e). The anomalous equatorward meridional wind during El Niño Modoki periods brings more ozone-poor air from the middle-high-latitude lower stratosphere to the tropical lower stratosphere in the longitude band of 60^°-120^°W. Figure 6f shows that the vertical velocity over the tropical eastern Pacific (60^°-120^°W) is stronger in El Niño Modoki events. The intensity of vertical velocity would bring more ozone-poor air from the upper troposphere to the stratosphere. The anomalous horizontal and vertical transports reduce ozone concentrations in the tropical lower stratosphere over the longitude band of 60^°-120^°W, causing a significant reduction in stratospheric column ozone over the tropical eastern Pacific (60^°-120^°W) during El Niño Modoki periods (Fig. 6b). The strongest meridional wind anomalies over this longitude range occur in the Southern Hemisphere, so that the lowest ozone concentrations in the tropical region are located over the tropical eastern Pacific in the Southern Hemisphere.

The radiation effect of tropospheric ozone has important impacts on climate (Liao et al., 2004, 2006; Liao and Seinfeld, 2005). ENSO activities dominate the variations of the Hadley cell (Quan et al., 2004), and have been shown to profoundly impact the Walker circulation by displacing areas of convective activity. These circulation changes influence the distribution of chemical constituents in the troposphere, including ozone (Chandra et al., 1998, 2009; Sudo and Takahashi, 2001; Chandra et al., 2002; Ziemke and Chandra, 2003; Zeng and Pyle, 2005; Doherty et al., 2006; Chandra et al., 2009; Lee et al., 2010; Randel and Thompson, 2011). Observations of TCO (Ziemke et al., 2010) and numerical simulations (Oman et al., 2011) indicate that the variations of ozone in the troposphere are dominated by the response to ENSO over the western and eastern Pacific. The impact of ENSO on tropospheric ozone has been extensively studied; however, changes in tropospheric ozone associated with El Niño Modoki have not yet been examined in detail.

Figure 7 shows latitude-height cross sections of composite ozone anomalies below 200 hPa associated with canonical El Niño (Fig. 7a) and El Niño Modoki (Fig. 7b). Canonical El Niño is associated with increased tropospheric ozone in the Southern Hemisphere. In the Northern Hemisphere, ozone is reduced in the lower troposphere but increased in the middle-upper troposphere (Fig. 7a). However, these effects of canonical El Niño events on tropospheric ozone are almost insignificant. El Niño Modoki is associated with a significant increase in ozone throughout the troposphere except for the north polar region, with a maximum increase in the middle troposphere at low latitudes.

Figure 7.  The same as Fig. 3, but for zonally-averaged tropospheric ozone anomalies composited for (a) canonical El Niño events and (b) El Niño Modoki events.

The EOF analysis of the 1980-2010 time series of stratospheric ozone presented in section 3 indicates that El Niño Modoki plays a key role in determining stratospheric ozone variability. A similar EOF analysis is performed on the time series of tropospheric ozone over the same period Figure 8 shows the spatial and temporal variations of the first and second EOF modes of tropospheric ozone. The spatial pattern of EOF1 which accounts for 50% of the total variance, corresponds to tropospheric ozone anomalies during El Niño Modoki events (Figs. 8a and 7b). The variation of PC1 is likewise strongly correlated with EMI (Fig. 8b), with a linear correlation coefficient of 0.57 EOF2 is also significant accounting for 22% of the variance (Fig. 8c). It is found that the variation of PC2 is significantly anti-correlated with the variation of lower stratospheric ozone (Fig. 8d), with a linear correlation coefficient of -0.54 This anticorrelation implies that the second EOF mode may be associated with NO_x-related chemistry in the atmosphere, which has opposite effects on tropospheric and stratospheric ozone (NO_x creates tropospheric ozone but depletes stratospheric ozone). This may be why we found that variations of PC2 are significantly anti-correlated with the variations of lower-stratospheric ozone Figure 8 also indicates that El Niño Modoki has been the most important factor in determining tropospheric ozone variability over the past 30 years. Figure 8e shows the variations of PC1 and averaged tropospheric ozone anomalies Changes in tropospheric ozone are significantly correlated with PC1, which is associated with El Niño Modoki This correlation implies that El Niño Modoki also plays a key role in controlling the variations of tropospheric ozone over the past three decades.

Figure 8.  The same as Fig. 4, but for zonally-averaged tropospheric ozone anomalies. The range for performing the EOF analysis is from 1000 hPa to 200 hPa and from 90^°S to 90^°N.

Previous studies have demonstrated that ENSO changes the longitudinal distribution of tropospheric ozone at lower latitudes (Ziemke et al., 2010; Oman et al., 2011). Figure 9 shows anomalies in tropospheric column ozone associated with both types of El Niño events. Tropospheric column ozone is calculated assuming that the troposphere extends from the tropopause to 1000 hPa, where the tropopause is defined as in section 3. These cross sections allow further comparison and analysis of tropospheric ozone anomalies associated with the two types of El Niño events. Due to the effect of canonical El Niño on convection and Walker circulation, the tropospheric ozone anomalies associated with canonical El Niño are positive in the western Pacific but negative anomalies occur in the middle and eastern Pacific (Fig. 9a). These results are consistent with previous studies (e.g., Chandra et al., 1998; Thompson and Hudson, 1999; Ziemke et al., 2010; Oman et al., 2011). By contrast, El Niño Modoki is associated with a significant increase in ozone throughout the tropical troposphere, with a maximum increase over the central-eastern Pacific (Fig. 9b). Figure 9 indicates that the tropospheric ozone anomalies associated with El Niño Modoki are completely different from those associated with canonical El Niño. Figure 9 also illustrates that El Niño Modoki induces a more profound increase in tropical tropospheric ozone than canonical El Niño. The increases in tropospheric ozone at low latitudes are transported poleward, which may lead to ozone increases at whole middle latitudes in the Southern Hemisphere and over the longitude ranges of 0^°-90^°E and 0^°-180^°W of middle latitudes in the Northern Hemisphere during El Niño Modoki events. The next step is to understand how El Niño Modoki causes significant increases in ozone in the tropical troposphere. The results also agree with the above if tropospheric column ozone is calculated assuming that the troposphere extends from 200 hPa to 1000 hPa.

Figure 9.  Latitude-longitude cross sections of tropospheric column ozone anomalies (%) composited from 1980-2010 ERA-Interim data for (a) canonical El Niño events and (b) El Niño Modoki events. Contour intervals are 2% for ozone anomalies. Solid lines represent positive values; dashed lines represent negative values. Tropospheric column ozone is calculated assuming that the troposphere extends from the tropopause to 1000 hPa. Shaded areas are significant at the 95% confidence level (Student's t-test). Latitude-longitude cross sections of tropical OLR anomalies composited from 1980-2010 NOAA data for (c) canonical El Niño events and (d) El Niño Modoki events. Contour intervals are 3 W m^-2. Latitude-longitude cross sections of tropospheric water vapor concentration anomalies averaged from 1000 hPa to 200 hPa composited from 1980-2010 ERA-Interim data for (e) canonical El Niño events and (f) El Niño Modoki events. Contour intervals are 1000 ppmv. Solid lines represent positive values; dashed lines represent negative values.

Tropospheric ozone concentrations are affected by stratosphere-troposphere exchange (STE), tropical convective activity, the amount of UV radiation reaching the troposphere, tropospheric water vapor concentration, and tropospheric chemistry. Previous studies have pointed out that changes in STE caused by ENSO significantly affect tropospheric ozone (Langford et al., 1998; James et al., 2003; Zeng and Pyle, 2005); however, STE primarily influences ozone in the extratropical troposphere. Changes in the horizontal distribution of tropical convection associated with the two types of El Niño are shown in Figs. 9c and d. Outgoing longwave radiation (OLR) data, represented as an index of convection, from 1980 to 2010 are obtained from the NOAA Climate Data Center (http://www.cdc.noaa.gov/). Due to the water vapor concentration substantially affecting the lifetime of ozone in the troposphere, the anomalies of tropospheric water vapor concentration averaged from 1000 hPa to 200 hPa during the two types of El Niño are also shown in Figs. 9e and f. Convective activity is strengthened in the eastern Pacific and weakened in the western Pacific during canonical El Niño events (Fig. 9c). The water vapor anomalies correspond to the convection anomalies (Figs. 9e and c). These changes in convective activity and water vapor are indeed associated with negative ozone anomalies in the eastern Pacific and positive ozone anomalies in the western Pacific (Figs. 9e, c and a). However, during El Niño Modoki periods, the water vapor anomalies are associated with convection anomalies, but those anomalies are not in accordance spatially with the ozone anomalies (Figs. 9f, d and b).

Ozone in the troposphere is also affected by UV radiation. Tropospheric ozone production by chemical processes is more efficient when solar radiation is more intense, and vice versa. Stratospheric column ozone changes can therefore affect tropospheric column ozone: decreases in stratospheric column ozone increase the amount of solar UV radiation reaching the troposphere, while increases in stratospheric column ozone decrease the amount of solar UV radiation. We found that tropospheric column ozone is significantly anti-correlated with stratospheric column ozone only in the tropics; the correlation coefficient is -0.67 (Fig. 10). The correlation coefficients between tropospheric column ozone changes and stratospheric column ozone changes in the northern and southern extratropics are -0.13 and 0.11, respectively. This strong anti-correlation implies that the intensity of UV radiation can influence tropospheric ozone changes in the tropics, and represents the possible mechanism of stratospheric ozone affecting tropospheric ozone during El Niño Modoki events. Certainly, the mechanism needs further verification.

Figure 10.  Normalized variations of stratospheric column ozone (solid line) and tropospheric column ozone (dashed line) anomalies averaged for (a) (25^°-90^°N, 180^°W-180^°E), (b) (25^°S-25^°N, 180^°W-180^°E), and (c) (25^°-90^°S, 180^°W-180^°E). The variations of stratospheric/tropospheric column ozone were obtained by removing the annual cycle and linear trend from the original time series.

The effects of the two types of El Niño events (canonical El Niño and El Niño Modoki) on global ozone from 1980 to 2010 have been investigated and compared using composite analyses of ERA-Interim reanalysis data and the N3I and EMI data. Canonical El Niño events lead to a positive ozone anomaly at high latitudes in the Northern Hemisphere and a negative ozone anomaly in the tropical lower stratosphere. The pattern of ozone anomalies associated with El Niño Modoki is totally different from those associated with canonical El Niño. This is because El Niño Modoki can more strongly enhance BD circulation compared to canonical El Niño, which transports more ozone-poor air to the lower stratosphere. Significant anomalies in the local meridional circulation over the longitude band of 60^°-120^°W in both hemispheres during El Niño Modoki periods generate a greater decrease in stratospheric column ozone over the tropical eastern Pacific than over other tropical areas.

El Niño Modoki events significantly increase the tropospheric column ozone, and the possible mechanism for this is changes in stratospheric column ozone affecting tropospheric column ozone in the tropics. Stratospheric column ozone decreases substantially during El Niño Modoki events in the tropics, increasing the intensity of solar UV radiation in the tropical troposphere. The process enhances the efficiency of chemical ozone production in the troposphere and induces a significant positive anomaly in tropospheric column ozone.

The EOF analysis of time series of stratospheric and tropospheric ozone anomalies over the past three decades reveals several key points. In the stratosphere, the leading EOF mode is associated with El Niño Modoki, which accounts for 27% of the total variance, while the second mode corresponds to the QBO, which accounts for 10% of the total variance. The leading EOF mode in the troposphere is also associated with El Niño Modoki, which accounts for 50% of the total variance. The second mode possibly corresponds to NO_x-related chemistry, which accounts for 22% of the total variance. These results indicate that El Niño Modoki has played a key role in controlling the variations of global ozone in the last three decades.