Interannual variability of summertime aerosol optical depth over East Asia during 2000–2011: a potential influence from El Niño Southern Oscillation

Aerosols degrade air quality, perturb atmospheric radiation, and impact regional and global climate. Due to the rapid increase in anthropogenic emissions, aerosol loading over East Asia (EA) is markedly higher than other industrialized regions, which motivates a need to characterize the evolution of aerosols and understand the associated drivers. Based on the MISR satellite data during 2000–2011, a wave-like interannual variation of summertime aerosol optical depth (SAOD) is observed over the highly populated North China Plain (NCP) in East Asia. Specifically, the peak-to-trough ratio of SAOD ranges from 1.4 to 1.6, with a period of 3–4 years. This variation pattern differs apparently from what has been seen in EA emissions, indicating a periodic change in regional climate pattern during the past decade. Investigations of meteorological fields over the region reveal that the high SAOD is generally associated with the enhanced Philippine Sea Anticyclone Anomaly (PSAA) which weakens southeasterlies over northeastern EA and depresses air ventilation. Alternatively, higher temperature and lower relative humidity are found to be coincident with reduced SAOD. The behavior of PSAA has been found previously to be modulated by the El Niño Southern Oscillations (ENSO), therefore ENSO could disturb the EA SAOD as well. Rather than changing coherently with the ENSO activity, the SAOD peaks over NCP are found to be accompanied by the rapid transition of El Niño warm to cold phases developed four months ahead. An index measuring the development of ENSO during January–April is able to capture the interannual variability of SAOD over NCP during 2000–2011. This finding indicates a need to integrate the large-scale periodic climate variability in the design of regional air quality policy.


Introduction
Ambient aerosols adversely impact human health (Janssen et al 2011, Pope et al 2002, Liu et al 2009a, reduce visibility (Park et al 2003, Langridge et al 2012 and perturb climate (Kaufman et al 2002, Ramanathan et al 2001. These negative effects raise a need to better constrain the factors which control the abundance of aerosols and its evolution from regional to global scale.
Aerosols originate from direct release of primary emissions or indirect formation through oxidation of precursor gases (e.g., SO 2 and VOCs, Liu et al 2009b). Large variability in aerosol concentrations may have significant consequences on regional climate and human health. Both scattering (e.g., sulfate, nitrate, and organics) and absorbing (e.g., dust, black and brown carbon) aerosols change atmospheric radiation, leading to a cooling/warming effect on global climate (Koch et al 2011). Besides, hydrophobic (e.g., dust, black carbon and part of primary organic matters) and hydrophilic (e.g., sulfate, nitrate, and secondary organics) aerosols initiate ice and liquid clouds (Prenni et al 2009, Ming et al 2007, and enhance the melting of land snow and sea ice cover (Mahowald et al 2011). These effects alter the earth albedo, and perturb global radiation budget. Moreover, particulate matter with aerodynamic diameter less than 2.5 µm (PM2.5) can penetrate deep into the lungs, causing cardiovascular disease, lung cancer, and premature death (Pope et al 2002, Liu et al 2009a.
In contrast to the long-lived species, such as CO 2 and CH 4 , aerosol concentrations vary significantly over space and time, mainly driven by the short lifetime of aerosols (∼1 week) (Liu et al 2009b) and the spatiotemporal inhomogeneity of emission sources (Lamarque et al 2010, Lin et al 2010. The distribution and evolution of global aerosol loading are observed continuously by satellite remote sensing as aerosol optical depth (AOD). The AOD data are widely used to understand the trend of emissions and the variability of air quality (e.g., Kaufman et al 2002, Lin et al 2010. While the trend in anthropogenic emissions plays an important role in shaping the AOD fluctuation, climate perturbations may affect AOD independently of changes in anthropogenic emissions through alterations of meteorological conditions and natural emissions. For instance, surface ozone is strongly correlated with temperature during summertime stagnant conditions (Jacob and Winner 2009). The export of Saharan dust and biomass burnings in tropics is found to be correlated with El Niño Southern Oscillation (ENSO) which is one of the prominent modes of large-scale climate variability (Hsu et al 2012).
However, over the populated regions, some researchers have observed a decreasing trend of AOD over the eastern US and Europe and a significant increasing trend of AOD over the eastern China and India (Augustine et al 2008, Hsu et al 2012, which mainly reflect a change in local anthropogenic emissions (Hsu et al 2012). For example in East Asia (EA), the bottom-up anthropogenic emissions have shown that SO 2 emissions (i.e., the precursors of sulfate aerosols) from China are tripled from 1980(peaked in 2006, Smith et al 2011. Other pollutants, such as NO x , black carbon (BC) and organic matter (OM), mostly show an increasing trend during the same period (Wang et al 2012). This rapid increase in anthropogenic emissions significantly enhances aerosol loading, and leads to severe air quality problems over China.
This study evaluates the interannual variability of AOD over EA and targets the underlying meteorological and climate drivers that impose on the fluctuation of AOD. Specifically, we focus on addressing the extent that the AOD variability over EA is tied to specific meteorological fields and large-scale climate perturbations like the ENSO. After a brief discussion of the data and method used in this study (section 2), we describe the seasonal and interannual variability of AOD over EA (AOD EA ), and examine the potential meteorological drivers based on correlation analysis (section 3). We then extend our exploration to the association of AOD EA to ENSO activities (section 4), and postulate the mechanism to explain this connection (section 5). Finally, we discuss the science and policy implications in section 6.

Data and methods
We acquired the multi-angle imaging spectroradiometer (MISR) (Diner et al 1998) level 3 AOD products at 558 nm (with horizontal resolution of 0.5 • × 0.5 • ) from the Atmospheric Sciences Data Center at NASA (http://eosweb. larc.nasa.gov). Compared to the ground based observations, MISR retrievals well depict the temporal aerosol trend in China with lowest biases in summer . Moreover, MISR data are more suitable than moderate resolution imaging Spectroradiometer (MODIS) retrievals in analyzing long-term and low-frequency variability, particularly over bright backgrounds (e.g., arid and semi-arid regions or large urban centers where high surface reflection and the lack of dense vegetation can bias MODIS retrievals) (Liu et al 2009c).
We analyze the MISR seasonal AOD data from 2000 to 2011, with a special focus on AOD above three subregions over the populated East Asia (e.g., the north China plain (NCP), southeastern China (SEC), and southwestern Japan (SWJ)). To evaluate the association of AOD to regional meteorological fields, we acquired the NCEP/NCAR reanalysis monthly data for the surface temperature, surface pressure, zonal and meridional wind, and relative humidity, which are provided by NOAA/OAR/ESRL PSD, Boulder, Colorado, USA (www.esrl.noaa.gov/psd/). In addition, we obtained monthly precipitation data from the Global Precipitation Climatology Project (GPCP, version 2) (Adler et al 2003).
To investigate the relationship between El Niño/La Niña events and AOD, we acquired the multivariate ENSO index (Wolter and Timlin 2011) (MEI) from www.esrl.noaa. gov/psd/enso/mei/ and the Niño 3.4 index (Barnston et al 1997) from http://climexp.knmi.nl/data/inino5.dat. Wolter and Timlin (2011) pointed out that the MEI index provides a more complete and flexible description of the ENSO phenomenon than single variable ENSO indices such as the Southern Oscillation index (SOI) or the Niño 3.4 index. Therefore, the MEI index is employed in this study to depict the variability of El Niño/La Niña episodes.
The association of interannual variability of AOD to meteorological fields and the ENSO index are examined based on correlation analysis. Correlation coefficients between two variables (e.g., X and Y) are calculated as: where cov(X, Y) represents the covariance between X and Y. σ X and σ Y are standard deviations of X and Y, respectively. Since the MISR archive starts in 2000, AOD data used in this study are limited to a 12-year period. Therefore, correlations less than 0.5 are not significant at the 95% confidence interval (CI). In order to quantitatively understand the interannual association between AOD EA and the ENSO activity, a linear regression is employed to explore the connection: where corr(X, Y) is the correlation coefficient between X and Y. The regression analysis indicates the change of AOD EA as a result of unit change of the ENSO index.

Seasonal and interannual variability of AOD over EA
As shown in figure 1, East Asia, particularly the NCP, is masked by thick aerosol cloud with seasonal AOD ranges from less than 0.4 in winter to more than 0.7 in summer. Unlike other regions such as Africa and the Amazon where high aerosol loadings originate from natural sources (i.e., dust or biomass burning), the high AOD over EA (AOD EA ) is mainly contributed by anthropogenic emissions of SO 2 , BC and organic aerosols (as well as the springtime dust particles transported from the deserts in northwestern China and Mongolia, see figure S1 in the supplementary materials available at stacks.iop.org/ERL/ 8/044034/mmedia for locations), which enhance ambient loadings of both primary and secondary aerosols (Tao et al 2012) and lead to frequent occurrence of severe haze over EA. Reddy et al (2005) estimated that sulfate, organic matter, and black carbon, respectively account for 60-80%, 5-10%, and 2-6% of total AOD EA . The AOD EA exhibits a distinct seasonal variation pattern (figure 1) that the lowest AOD EA appears in December-January-February (DJF, i.e., the boreal winter) and the highest shows in June-July-August (JJA, i.e., the boreal summer). During boreal winter, frequent cold fronts sweep the pollution away and replace it with clean polar air. In addition, formation of secondary aerosols (i.e., sulfate and SOA) becomes slower due to a lack of oxidants and precursor gases , He et al 2013. In spring, prevailing winds (see figure S2 in the supplementary During the boreal summer, the prevailing East Asian monsoon over southern China dilutes and scavenges air pollutants, resulting in a decreased summertime AOD (SAOD) (Zhu et al 2012). In contrast, over northeastern EA particularly the NCP, the SAOD climbs to its maximum for a year due to a number of reasons such as enhanced production of secondary aerosols and increased moisture , Fu et al 2012, Heald et al 2010. After summer, AOD EA falls back to a level analogous to that in spring, but prominent AOD only stands over the eastern China. Figure 1 depicts a relative maximum of AOD over the NCP in JJA, coincident with the most populated region in China. When averaging the SAOD over the NCP (30 • -40 • N, 110 • -120 • E, i.e., SAOD NCP ), a wave-like interannual variation of the SAOD NCP is observed (figure 2). The SAOD NCP peaks are stable at ∼0.65 with a period of 3-4 years, exceeding the troughs (∼0.45) by 40-60%. Over southeastern China (20 • -30 • N, 110 • -120 • E), the fluctuation of SAOD is similar to that in NCP, but with a much weaker amplitude and an opposite phase, probably caused by a north-south shift of East Asian monsoon. Over Japan (45 • -55 • N, 140 • -150 • E), the variability of SAOD is different to either NCP or SEC, particularly after 2006.
The changes in local emissions or meteorological patterns could influence aerosols abundance and light extinction, resulting in SAOD fluctuation. As shown in Reddy et al (2005), the SAOD over EA is mostly anthropogenic origin. However, the wave-like oscillation of SAOD NCP is seemingly different to either the inverted U-shaped trend ( AOD measures the integrated extinction of aerosols on the atmospheric column. Its magnitude is determined by both aerosol column number density and light extinction efficiency. Aerosol types and their mixing states (i.e., internally or externally mixed) determine the refractive index and thereby the extinction efficiency of aerosols. Aerosol column abundance modulates the cumulative extinction of solar radiation. As a result, meteorological factors could influence AOD via either aerosol abundance or its extinction ability. Temperature, wind/pressure, and precipitation could influence the formation, transport and removal of aerosols, which alter aerosols spatiotemporal distribution. Relative humidity (RH) controls the uptake of water vapor by hydrophilic aerosols, directly influencing aerosol extinction efficiency.
To understand the potential effect of changing meteorological fields on the interannual variability of SAOD NCP , we conducted a correlation analysis over the East Asian domain. Figure 3 shows the correlation coefficients between SAOD NCP and different meteorological factors (i.e., surface temperature, RH, precipitation and SLP/wind fields) in each reanalysis grid during 2000-2011. As shown in figure 3(a), a weak anti-correlation pattern between SAOD NCP and surface temperature resides over the Yellow Sea, opposed to a weak positive association spreading over the southeastern China. This correlation pattern indicates that a lower temperature to the east of NCP or a higher temperature to the south of NCP may be associated with higher SAOD NCP . Correlation alone does not stand for the cause-effect relationship. Therefore, lower temperature could be either the cause of (e.g., lower temperature is usually associated with higher RH, which increases aerosols water uptake and light extinction, and enhances multiphase formation of semi-volatile organic species (Liu et al 2012)) or caused by (e.g., sometimes dense fog or scattering aerosols may block sunlight and cool the surface air) higher AOD. Therefore, to understand the mechanism which fundamentally determines the dynamical feedbacks between AOD and surface temperature over NCP, more analysis on the detailed chemical, physical and dynamical processes based on the fully coupled global climate model is needed, and will be examined in the follow-up studies. Figure 3(b) shows that the relationship between AOD NCP and RH follows a similar pattern to that of surface temperature depicted in figure 3(a) but with a reversed signal. In particular, the RH over the Yellow Sea and south Japan is positively correlated with SAOD NCP . Higher RH could enhance the uptake of water for water-soluble aerosols (e.g., sulfate, nitrate and certain organic species) and therefore increase aerosols extinction coefficient. In contrast, precipitation in general is an indicator measuring the removal of aerosols. However, precipitation itself is influenced by the abundance of aerosols. Aerosol particles may act as cloud condensation nuclei (CCN) and ice nuclei (IN), which modify cloud properties (e.g. cloud droplet size and lifetime) and precipitation (e.g., Denman et al 2007). As inferred from figure 3(c), there is no direct linkage between precipitation and the fluctuation of SAOD NCP during the past decade. Figure 3(d) shows the correlations of SAOD NCP to the surface pressure (i.e., colors) and 850 mb wind field (i.e., arrows). SAOD NCP is strongly associated with a large-scale anticyclonic circulation anomaly over the western equatorial Pacific, and positively correlated with surface pressure centered at the Philippine Sea. This correlation indicates that SAOD NCP is linked to atmospheric circulation pattern over the western subtropical Pacific. Figure S2 (available at stacks.iop.org/ERL/8/044034/mmedia) depicts the summer mean surface pressure and wind vectors averaged from 2000 to 2011 over EA. Southeastern China is influenced by the air mass from South China Sea, while the air quality over NCP is mostly influenced by the air mass transported from the Yellow Sea. The intrusion of clean maritime inflows from western Pacific significantly dilutes aerosol pollution and decreases the SAOD NCP . Figure 3(d) also reveals that an intensified Philippine Sea subtropical high would trigger an anticyclone anomaly circulated over the eastern EA, which would block eastward transport of clean air from western Pacific and foster stagnation conditions favoring air pollution episodes over NCP. Therefore, the fluctuation of SAOD NCP is associated to the climate variability over western Pacific. Zhu et al (2012) simulated the surface PM2.5 concentrations over EA based on the global chemical transport model GEOS-CHEM driven with the assimilated meteorological fields. By comparing to the change of East Asian summer monsoon index, they indicated that the increases in surface aerosol concentrations over eastern China are contributed by decadal-scale weakening of the East Asian summer monsoon.

Potential influence by the El Niño episodes
ENSO is the most prominent coupled atmosphere-ocean phenomenon that causes seasonal and interannual variability of global climate (Wolter and Timlin 2011). ENSO is also believed to impact precipitation over East China Hu 2004, Xue and. As indicated by Lau and Nath (2006), many ENSO-related circulation changes in East Asia are connected to the evolution of Philippine Sea anticyclone anomalies (PSAA). Therefore, the linkage of SAOD NCP to PSAA elicits a hypothesis that the variability of SAOD NCP is influenced considerably by the El Niño episodes.
However, the interannual variability of SAOD NCP is not directly linked to the amplitude of ENSO episodes. Figure 4 shows that the variability of SAOD NCP is inconsistent with the time series of multivariable El Niño index (MEI) averaged in JJA. Results are similar for the Niño 3.4 index. As indicated by Wang et al (2009), summer monsoon over the South China   Sea is mainly affected by the decaying phase of ENSO after 1970s. The ENSO's influence on summer monsoon infers that rather than the strength of El Niño episodes, the rate of ENSO change could play an important role in affecting SAOD NCP anomalies. Table 1 lists the correlation coefficients between the SAOD NCP and the rate of change in the MEI index in different seasons. Correlation is highest during the winter-spring season, specifically in a period from January to April (represented as MEI A-J ), indicating that a rapid transition from El Niño cold to warm phase during January to April is generally followed by a suppressed SAOD NCP (see the anti-correlation pattern in figure 4). In general, years with lower MEI A-J (here 'lower MEI A-J ' means more negative MEI A-J values) are associated with higher SAOD NCP , and vice versa. It suggests that the ambient air quality and visibility over the most populated region in EA could be markedly influenced by the sea surface temperature (SST) anomalies over the equatorial eastern Pacific.

Mechanism driving the interannual variability of SAOD NCP
The role of remote ENSO forcing from the equatorial central Pacific in setting up the subsequent large-scale circulation pattern in controlling the summertime air pollution and AOD loadings over the NCP could be partially explained by the association between the MEI A-J index and development of tropical equatorial Pacific meteorological fields (i.e., the SLP, wind fields and surface temperature) from the boreal winter to summer (figure 5). A positive MEI A-J index in general depicts a rapid SST increase over the eastern equatorial Pacific during the transition period from winter to spring (see figure  S3 in the supplementary materials available at stacks.iop.org/ ERL/8/044034/mmedia). During JFM, a positive MEI A-J index is accompanied by an above-normal high pressure anomaly (figure 5 or below-normal low temperature anomaly, figure S3 available at stacks.iop.org/ERL/8/044034/mmedia) over the eastern equatorial Pacific, characterized mainly by a La Niña signal. In spring, a high MEI A-J index is generally subject to the retreat of high pressure anomaly and an enhancement of SST over the equatorial Pacific. This high MEI A-J index potentially generates a heat source which forces a cyclone anomaly over the North Pacific from mid-spring to summer (Lau and Nath 2006). Results from the stationary wave model calculation confirmed such a low pressure anomaly inferred from the meteorological difference between prominent El Niño and La Niña episodes (Lau and Nath 2006). Besides the low pressure anomaly at the mid-latitude Pacific, two more low pressure anomaly centers emerge at the eastern and western equatorial pacific (figures 5(d)-(e)) during years with high MEI A-J index. The one residing over the western subtropical Pacific generates a cyclonic wind vector anomaly at 850 hPa, which intensifies the maritime inflow over the NCP, significantly ameliorating the air quality during high MEI A-J summers.
This ENSO-induced meteorological development feature persists even beyond the period 2000-2011. As shown in figure S4 in the supplementary materials (available at stacks.iop.org/ERL/8/044034/mmedia), a rapid westward displacement of the North Pacific low pressure anomaly from April to August is observed coincident with high MEI A-J during 1980-2011. This coincidence indicates that summertime air quality and radiative forcing over NCP could be inherently modulated by the pace of springtime transition of El Niño cold/warm phases.

Discussions
AOD measures the aerosol column extinction which is affected by both emissions and climate perturbations. Quantitatively decomposing the change of SAOD NCP into these two components relies on detailed process analysis in the state-of-the-science climate modeling system as well as accurate estimation of spatiotemporal variability of regional emissions. While the oscillation of SAOD NCP is found to be relevant to the ENSO activities, emission changes, triggered by economic growth/recession, urbanization, and implementation of emission control regulations (not fully captured by emission inventories), could play a significant role in controlling the oscillation of SAOD NCP (Wang et al 2012, Xu et al 2006. Strictly establishing the relationship between emission change and AOD variability is beyond the scope of this study, and will be investigated in future works. This study reveals that the intra-decadal variability of SAOD NCP could be modulated by ENSO activities even with the possible influence from emission fluctuation. Understanding the relationship between SAOD NCP and ENSO has significant policy inferences. A practical application is to empirically estimate the summertime air quality and visibility over one of the world most populated regions based on the development of winter-spring El Niño activities. When SST over the eastern equatorial Pacific experiences a rapid increase (decrease) in JFMA, the associated weakening (strengthening) of PSAA and enhanced (depressed) easterly over the Yellow Sea ameliorate (degrade) air quality or visibility over NCP in JJA. Therefore, it is of great necessity to consider the influence of large-scale periodic climate variability when crafting local air quality policies and tracking the subsequent progress.
Besides, the high loading of aerosols over East Asia is shown to account for the surface cooling and precipitation decrease over NCP during the past decades (Zhao et al 2006), and thus has a potential to influence regional climate. Climate perturbations in different places are linked through teleconnection. Alternations of El Niño/La Niña episodes set up a large-scale natural forcing (Kim et al 2012), which significantly alters the anthropogenic aerosol forcing over EA. Climate response to this forcing perturbation will be further propagated to other places via teleconnection. Therefore, it is important to understand the role of teleconnection in setting up the passage of a regional forcing to a much broader scale, and to explore the circumstances that the superimposed anthropogenic influence would coincide with or trigger extreme weather events.
Finally, the impact of MEI A-J on SAOD is not limited to the NCP. As shown in figure 6, strong correlation between MEI A-J and AOD over the west Africa in DJF, northeastern Asia in MAM and JJA, and the Amazon in SON, implies widespread ENSO-induced forcing of aerosols originating in tropical and boreal forests, and from biomass burning and anthropogenic activities. Signals over the remote oceans may also imply that the MEI A-J index is associated with long-range transport of air pollution (Lang et al 2008). Therefore, future studies on chemistry-climate interactions, long-range aircraft measurements as well as emission inventory development referenced on AOD may consider this El Niño effect.