Shuwei Xi , Zhicong Yin , b , c , * , Huijun Wng , b , c

a Key Laboratory of Meteorological Disaster, Ministry of Education/Joint International Research Laboratory of Climate and Environment Change (ILCEC)/Collaborative Innovation Center on Forecast and Evaluation of Meteorological Disasters (CIC-FEMD), Nanjing University of Information Science & Technology, Nanjing, China

b Southern Marine Science and Engineering Guangdong Laboratory (Zhuhai), Zhuhai, China

c Nansen-Zhu International Research Center, Institute of Atmospheric Physics, Chinese Academy of Sciences, Beijing, China

Keywords:Sea surface temperature Tropical indian ocean Haze pollution CESM-LE

ABSTRACT Haze events in the Yangtze River Delta (YRD) region have recently been occurring more frequently and with dramatic damages inflicted on human and ecosystem health. In this study, observational analyses and numerical experiments are used to investigate the meteorological conditions associated with haze pollution, with the main emphasis on the impacts of the preceding sea surface temperature (SST) in the tropical Indian Ocean (TIO). The results show that the December SST in the TIO has a significant positive correlation with the number of haze days in January over the YRD, especially during 1999—2017. In December, the positive SST anomalies in the TIO heat the overlying air, and then in the following January provoke a Matsuno—Gill-like pattern and a series of Rossby wave—like trains in the upper troposphere, transmitting signals to the YRD and downstream through the Sea of Japan and Aleutian Islands. The cyclonic anomalies in the YRD seem to significantly weaken the East Asian jet stream by means of anomalous easterlies, and subsequently affect the climate in the region. Near the surface, the increased surface air temperature and southerly winds, along with the decreased surface wind speed,accompanied by influences from upstream areas, are conducive to the occurrence of haze. These observational results were also reproduced well in CESM-LE simulations.

1. Introduction

Haze pollution has recently been occurring more frequently in China,particularly in the North China Plain, Fenwei Plain, Yangtze River Delta(YRD), Pearl River Delta, and Sichuan Basin regions ( Wang, 2018 ;An et al., 2019 ). According to recent research, anthropogenic aerosol emissions, closely associated with industrialization, urbanization, and agricultural events within the global economy, play a key role in affecting the development of haze events ( An et al., 2019 ). In addition, the air pollution in eastern China is partially exacerbated by global warming, which results in a stable atmosphere and weakening of cold air activity ( Wang et al., 2015 ; Wang and Chen, 2016 ; Cai et al., 2017 ). In terms of external and preceding climate drivers, the decline of Arctic sea ice could intensify winter haze pollution in eastern China ( Wang et al.,2015 ). The negative sea surface temperature (SST) anomalies over the subtropical Pacific intensify winter haze days over the north-central North China Plain ( Yin and Wang, 2016 ). Large-scale SST patterns such as the El Ni?o—Southern Oscillation (ENSO) and the Pacific Decadal Oscillation also have close connections with haze frequency in eastern China ( Gao and Li, 2015 ; Li et al., 2017 ). Meanwhile, the changes in meteorological conditions related to external forcing factors are of vital importance in the formation of winter haze, such as strong thermal inversion, calm and static surface wind, low planetary boundary layer height (PBLH), and high relative humidity, tend to accumulate pollutants within the source regions ( Yin et al., 2015 ; Gao and Chen, 2017 ;Yin and Wang, 2017 ; Yin et al., 2017 ; Che et al., 2019 ; Chen et al.,2019 ).

The above studies were mainly confined to analyzing the effects on haze pollution in the Northern Hemisphere. However, in the tropics, Indian Ocean SST is an important driver, which could influence the atmospheric circulations in East Asia ( Yang et al., 2007 , 2010 ). In higher latitudes, the impacts on precipitation in the Northern Hemisphere from the Antarctic Oscillation (AAO), provide a possible mechanism to explain the transmission of signals from the Southern Hemisphere ( Wang and Fan, 2005 ; Sun et al., 2009 ; Liu et al., 2015 ). Fan and Wang (2004) illustrated the linkage between the AAO and dust weather frequency in North China through meridional teleconnection and regional circulation over the Pacific Ocean. Zhang et al. (2019) demonstrated the influence of the AAO on haze pollution by inducing an SST warming tendency in the southern Indian Ocean. However, far fewer studies have focused on haze pollution in the YRD, defined as the foremost economic zone in China, as well as the relevant changes in climate, especially in the Southern Hemisphere. Therefore, the main aims of this study are to examine the impacts of SST in the tropical Indian Ocean (TIO) on haze pollution in the YRD and to investigate the possible physical mechanism involved. The findings will help us to better understand winter haze pollution over the YRD, as well as the remote influences of the TIO SST.

2. Data and methods

Monthly SST data for the period 1979—2017 were downloaded from the Met-Office Hadley Center, with a horizontal resolution of 1°×1°( Rayner et al., 2003 ). The 2.5°×2.5°monthly reanalysis data used here were from the National Centers for Environmental Prediction—National Center for Atmospheric Research ( Kalnay et al., 1996 ) and included the zonal and meridional wind, vertical velocity, relative humidity, air temperature at different pressure levels, and surface air temperature (SAT).Furthermore, the 1°×1° planetary boundary layer height (PBLH) was derived from the ERA-Interim dataset ( Dee et al., 2011 ). The Ni?o-3.4 index was from the Climate Prediction Center.

The number of haze days used here was calculated in accordance with Yin et al. (2017) . A haze day was defined as a time when the visibility decreased to 10 km and the relative humidity was lower than 90%, after excluding other weather phenomena affecting visibility (i.e., precipitation, dust, sandstorms, and blowing snow). The monthly total emissions, by pollutant, were derived from the Peking University emissions inventory, with a horizontal resolution of 1°×1°,from 1979 to 2013. The pollutants included black carbon, NH, NO,organic carbon, SO 2 , PM 10 , and PM 2.5 , and the emission sources of these pollutants involved energy production, industry, transportation,residential and commercial activity, deforestation and wildfire, and agriculture.

Simulations from the Community Earth System Model Large Ensemble (CESM-LE) datasets were employed ( Kay et al., 2015 ), with a horizontal resolution of 0.9°latitude ×1.25°longitude and 30 vertical levels.The CESM-LE simulations were completed by the fully coupled CESM model, which combined the historical simulation during 1979—2005 and the data during 2006—17 from the Representative Concentration Pathway 8.5 forcing simulation. The variables used here included SST,meridional and zonal winds, and SAT.

The term JHDrefers to the mean number of January haze days over the YRD (28°—34°N, 118.5°—122.5°E), with similar notation —JHDand JHD—for the North China Plain (34°—43°N,114°—120°E) and Pearl River Delta (22°—25°N, 110°—115°E), respectively. The SSTaindex is defined as the averaged SST over the TIO(25°S—10°N, 40°—100°E). In this study, the analyses are based on calculations after removal of the linear trend. To facilitate a convenient understanding of the data, years are labeled based on the four seasons.For example, the year 1979 defined from March 1979 to February 1980 is denoted as 1979.

Fig. 1. The CCs between JHDYRD and (a) a cross section (118.5°—122.5°E mean)of relative humidity (rhum, shading), air temperature (air, contours), and wind(arrows), (b) surface wind speed (wspd, shading) and surface wind (arrows),and (c) PBLH (shading) and SAT (contours), in January from 1979 to 2017.The black dots indicate that the CCs exceed the 95% confidence level ( t -test).The green box represents the YRD. The solid contours indicate that values are positive; dashed contours indicate values are negative; and the contour interval is 0.1. The linear trend has been removed.

3. Results

From 1979 onward, JHDincreases remarkably, and more than half. JHD YRD values exceed those of JHD NCP during 1993—2015, as well as JHDbasically from 1979 (Fig. S1). The correlation coefficient(CC) between JHD YRD and JHD NCP is 0.15, and that with JHD PRD is 0.28, both of which hardly exceed the 95% confidence level, implying that haze pollution in the YRD is serious and worth studying. Previous studies have demonstrated the important role of anthropogenic emissions ( An et al., 2019 ). However, the CC decreases from 0.55 to ? 0.21 after removing the linear trend (Fig. S2(a, b)), meaning that the influence of anthropogenic emissions weakens from the perspective of interannual variation. Therefore, it is necessary to analyze the features of haze pollution in the YRD and its associated climate drivers. However, due to the lack of previous studies on the atmospheric circulations associated with JHD YRD , the CCs between JHD YRD and local meteorological conditions are shown in Fig. 1 . When JHDis more than normal,there are anomalous secondary circulations over the YRD and its north.The ascending branch weakens meridional circulations, strengthening the development of severe haze over the Beijing—Tianjin—Hebei region( Zhong et al., 2019 ). The sinking motions inhibit convective activity,and together with the negative relative humidity anomalies from the middle and upper troposphere to the ground ( Fig. 1 (a)) lead to weakened wet removal in the YRD, which is not apparent in the NCP. When reaching the ground, the vertical circulation causes surface southerly winds and decreases the wind speed ( Fig. 1 (b)). The southerly anomalies weaken the prevailing northerlies, reducing the invasion of cold air and providing warm air and static winds for haze formation, which are not conducive to haze dissipation. The positive air temperature anomalies from the middle troposphere result in the difficulty in exchanging thermal energy between the YRD and its north, and thus the YRD could remain warm for a long time ( Fig. 1 (a)). Moreover, the SAT increases, and the PBLH declines significantly in the northwest of the YRD ( Fig. 1 (c)),limiting the upward dispersion of the pollutant particles.

As shown in Fig. 2 (b), the December SST in the TIO (25°S—10°N,40°—100°E) and the JHDare positively correlated, with a CC of 0.38,above the 95% confidence level, meaning that preceding increased SST would probably intensify haze pollution over the YRD. Moreover, during P2 (1999—2017), the CC is 0.6 ( Fig. 2 (a)), above the 99% confidence level, which is more significant than that during P1 (1979—98), i.e., only 0.06; thus, the following analyses are all based on P2. Previous studies have shown that ENSO could influence Indian Ocean SST ( Yang et al.,2007 ) and modulate the interannual variability of winter haze days over eastern China ( He et al., 2019 ). After removing the influences of ENSO(measured by the annual mean Ni?o3.4 SST index) on SST in the TIO, the positive CC with JHDis still present and increases slightly, i.e., 0.61,indicating that the connection exists regardless of the remote oceanic forcing, especially the effects of ENSO.

Fig. 2. (a) The CCs between the December SST and JHDYRD from 1999 to 2017.The black dots indicate that the CCs exceed the 95% confidence level ( t -test).The black box highlights the significantly correlated areas, which were used to calculate the SST anomaly indices. It represents the TIO. The linear trend has been removed. (b) The variations in the standardized SSTaTIO (orange solid line)in December and JHDYRD (blue bars) from 1979 to 2017 after detrending .

Therefore, how do anomalous signals of the SST transmit to the Northern Hemisphere and subsequently affect the January weather conditions over the YRD? Positive SSTa TIO leads to diabatic heating to some extent, provoking a Matsuno—Gill-like pattern ( Matsuno, 1966 ;Gill, 1980 ). The disturbances of the anticyclone over the TIO likely cause a series of Rossby waves propagating to the YRD and then spreading downstream through the Sea of Japan and Aleutian Islands. In addition, for the cyclonic anomalies centered in the YRD, zonal easterlies prevail in the northern YRD, largely weakening the East Asian jet stream ( Fig. 3 (a)). Therefore, the ascending motions and southward cold air tends to be suppressed, providing favorable conditions for the haze weather. From the local meteorological conditions, the secondary circulation is forced over the YRD ( Fig. 3 (b)). The sinking motions, along with the dry air, though weak, are consistent with the phase in Fig. 1 (a).Near the surface, southerlies prevail, and wind speeds are significantly reduced. The warm air temperature anomalies tend to weaken the East

Fig. 3. The CCs between the December SSTaTIO and (a) zonal wind (shading)and wind (arrows) at 200 hPa, (b) a cross section (118.5°—122.5°E mean) of relative humidity (rhum, shading), air temperature (air, contours), and wind(arrows), and (c) surface wind speed (wspd, shading), SAT (contours), and surface wind (arrows), in January from 1999 to 2017. The solid contours indicate values are positive; dashed contours indicate values are negative; and the contour interval is 0.1. The white dots indicate that the CCs exceed the 95% confidence level ( t -test). The black (green) box represents the TIO (YRD). The letters“C ”and “A ”represent the centers of the cyclonic and anticyclonic circulations,respectively. The linear trend has been removed.

Fig. 4. Composite differences in (a) SST in December, (b) 200-hPa wind in January, and (c) 925-hPa wind speed (shading) and wind (arrows) in January, (d) SAT(shading) in January. The black (green) box represents the TIO (YRD). The results are based on 40 ensembles of CESM-LE simulations. The gray shading and black dots indicate that the composite differences exceed the 95% confidence level ( t -test).

Next, we designed a numerical experiment using CESM-LE data, with 40 available members. Given that the relationship was stable over time,the variables from 1979 to 2017 were ultimately employed. To confirm the above results, the years when the SST anomalies concentrated in the TIO were selected, and the differences between the positive and negative SSTawere verified as responses to the SST anomalies. In Fig. 4 (a),during December, the SST anomalies are apparent in the critical region.In the following January, according to the corresponding atmospheric circulations and meteorological conditions, the composite differences are consistent with the observed results, although there are a few differences in the intensities and locations. The Rossby wave—like trains are reproduced well in the model, in which the cyclonic anomalies are clear over the YRD at 200 hPa ( Fig. 4 (b)). In the lower troposphere, southerly wind prevails over the YRD, weakening cold air from high latitudes and decreasing wind speeds significantly ( Fig. 4 (c)). Furthermore, warmer air conditions are evident in the YRD ( Fig. 4 (d)). The consistent results in numerical experiments verify that increased SST in the TIO would provide favorable conditions to intensify haze pollution over the YRD.Asian winter monsoon, and hence cold air is inhibited ( Fig. 3 (c)). All of these factors contribute greatly to aggravating haze pollution in the YRD. In addition, with the help of strong northwesterly winds, more haze particles are blown to its downstream regions. The northerlies and southerlies converge near the YRD, which is conducive to the accumulation of pollutants ( Fig. 3 (c)). Under the circumstances of these favorable conditions, haze could be exacerbated rapidly and perniciously over the YRD.

4. Conclusions and discussion

In this study, we investigated the close relationship between December SSTaand JHD. The significant CC of 0.6 indicated that preceding positive SST in the TIO could intensify January haze pollution over the YRD. As indicated above, SSTaproduces a Matsuno—Gilllike pattern in the TIO at 200 hPa, and hence a series of Rossby waves are induced one by one to transmit the signals to downstream regions.Specifically, they propagate to the YRD and then through the Sea of Japan and Aleutian Islands. In particular, the cyclonic anomalies in the YRD seem to significantly weaken the East Asian jet stream by means of anomalous easterlies over the north of the YRD. The SSTa TIO not only affects the atmospheric circulation but also influences the local meteorological conditions associated with haze, particularly increasing SAT,bringing southerly wind, and reducing the surface wind speed over the YRD. Moreover, wind convergence plays an important role in aggravating haze pollution, bringing more pollutants and leading to the accumulation of haze particles near the YRD. The linkages and relevant physical mechanisms were well reproduced via CESM-LE simulations, further affirming the causality. When considering P1, the CCs were no longer evident, not only in the atmospheric circulations, but also in the local meteorological conditions, which were non-significant or even reversed in the YRD (figures omitted).

Previous studies have revealed that, by means of the atmosphere and SST, the Northern and Southern Hemispheres could interact frequently( Jones et al., 2018 ; Lu and Guan, 2009 ). The findings of this research will help us to better understand the remote influences of the TIO SST on haze pollution and provide a new possible bridge connecting the Northern and Southern Hemispheres and transmitting signals, so further studies are still needed for us to understand the TIO well.

Declaration of Competing Interest

The authors declare no conflicts of interest.

Funding

The National Key Research and Development Plan [grant number 2016YFA0600703 ], the National Natural Science Foundation of China[grant numbers 41705058 , 91744311 , and 41991283 ], and funding from the Jiangsu Innovation and Entrepreneurship team, supported this research.

Supplementary materials

Supplementary material associated with this article can be found, in the online version, at doi: 10.1016/j.aosl.2021.100042 .