Understanding severe winter haze events in the North China Plain in 2014 : roles of climate anomalies

Atmospheric pollution has become a serious environmental and social problem in China. Over the past 30 years, the number of winter (December–February) haze days over the North China Plain (WHDNCP) was greatest in 2014. In addition to anthropogenic influence, climate anomalies also played a role. Thus, it is necessary to analyze the anomalous atmosphere circulations associated with haze pollution of this year in detail. Near the surface, the weaker East Asian winter monsoon pattern, causing southerly winds over the North China Plain, could aggravate the situation of haze. In the lower and middle troposphere, taking the anticyclone circulation over North China as an intermediate system, the positive phases of the eastern Atlantic/western Russia (EA/WR), the western Pacific (WP), and the Eurasia (EU) patterns led to a worse air pollution dispersion condition that contributed to a larger number of WHDNCP. In 2014, these three patterns could be recognized from the wind anomalies in the lower troposphere. The preceding autumn (September–November) Arctic sea ice (ASI) anomalies over the eastern Hemisphere and the warmer winter surface over Eurasia might have induced or intensified the positive EA/WR pattern in 2014. These two external forcings, together with the pre-autumn sea surface temperature anomalies in the Pacific, might have also stimulated or enhanced the positive EU-like patterns. The anomalous surface temperature in autumn 2014 was efficient in intensifying anomalous circulations such as the positive phase of the WP pattern. The opposite case of minimum WHDNCP in 2010 further supports the mechanism of how EA/WR and WP patterns and associated external factors altered the local climate conditions to impact the WHDNCP.


Introduction
Related to booming economic development, atmospheric pollution has become a serious environmental and social problem in China (Ding et al. 2014;Wang et al. 2016). Particularly after the persistent heavy fog and haze events in January 2013, haze pollution has become more severe (Zhang et al. 2014;Zhao et al. 2014;Li et al. 2015) and has presented certain negative effects on human health (Yin et al. 2011;Chen et al. 2013). The North China Plain (NCP), a location in which the population density is quite high, was one of three haze-prone regions in China. The winter (December-February) haze 30 pollution over NCP (34-43 o N, 114-120 o E) in 2013 and 2014 was the most serious of these events in the past 30 years (Yin et al. 2015a). Therefore, the objective of this study was to examine the related climate conditions (i.e., atmospheric circulation anomalies and external forcings) that were responsible for the extreme haze events in 2014.
There is no doubt that the anthropogenic emissions were the fundamental cause for the long-term variation of haze days in eastern China (Wang et al. 2013). The enormous energy consumption supplied enough particulates, so the atmosphere tended 35 to reach saturation. Thus, the impact of meteorological conditions is highlighted and the climate conditions are also vital contributors to the interannual variation of haze (Liao et al. 2014). For example, the joint effect of fast increase of total energy consumption, rapid decline of Arctic sea ice (ASI) extent and reduced precipitation and surface winds intensified the haze pollution in central North China after 2000 (Wang et al. 2016). Early studies documented that the East Asian winter monsoon (EAWM) had weakened after 1986, which led to an increase of winter haze days (WHD) over NCP (WHD NCP ) 40 (Yin et al. 2015a;Yin et al. 2015b;Li et al. 2015). The decline of the preceding autumn (September-November) ASI from 1979 to 2012 greatly intensified haze pollution in eastern China, the variance contribution of which was 45-67% ). Sea surface temperature (SST) over the subtropical western Pacific (SWP) showed significantly negative correlation with WHD NCP . SWP-SST weakened EAWM circulation, leading to a favorable environment for haze with stable atmosphere and potential for hygroscopic growth (Yin et al. 2016a).
high-latitude polar region. The EU positive phase showed negative centers over the polar region (70-80 o N, 60-90 o E) and the Japan Sea (35-45 o N, 120-140 o E), and a positive center over , which 50 accounted for the severe drought in 2014 ). Another Eurasian teleconnection, known as the EA/WR pattern (Barnston et al. 1987), was composed of negative centers over Central-North Atlantic and to the north of the Caspian Sea, and positive centers over Europe and North China. The positive phase of these two continental Rossby wave trains might have led to significant warming over the northern portion of Eastern Asia , indicating weaker cold air.
Therefore, we speculated that external forcings such as the SST, ASI and land surface temperature (TS) might impact 55 teleconnection patterns in the atmosphere and then the teleconnection patterns could alter the local climate anomalies to modulate the WHD NCP remotely. Climate research on haze pollution in China is quite a new endeavor but is still insufficient, especially with respect to investigation into the mechanism that causes extreme haze events. Thus, the roles of climate anomalies in the winter haze in 2014 over NCP were investigated in this study and were expected to improve prediction skill for WHD NCP .

60
The remainder of this paper is organized as follows. The data and methods are clearly explained in Section 2. The climatic reality of severe WHD NCP in 2014 and associated atmospheric circulations are analyzed in Section 3. Section 4 describes investigation of the physical mechanism using the singular value decomposition (SVD) technique. Brief conclusions and selected discussions are presented in Section 5.

Datasets and methods
composition at this location was chosen to best reflect the natural state or rather the background conditions of the atmosphere (Yao et al. 2012). The PM 2.5 data were monitored every 5 minutes using two methods: tapered element oscillating microbalance (TEOM) and β-ray (since 2013).
In China, the temporal range and quality of the observed atmospheric compositions cannot support climatic haze research. To demonstrate the representation of haze data reconstructed primarily by visibility, the hourly visibility and concentration of 90 the atmospheric composition were shown, and the correlation coefficients were computed ( Figure S1). In addition to significantly positive correlation with O 3 , the correlation coefficients between visibility and seven compositions were all negative and exceeded the 99.99% confidence level in winter 2014. PM 2.5 was the main reason for haze pollution, and the correlation coefficients with visibility in Beijing were −0.51 (Baolian) and −0.48 (Shangdianzi). When visibility was less than 7.5 km and 1 km, the mean PM 2.5 mass concentrations at Baolian were approximately greater than 100 and 200μg/m 3 , 5 respectively. Thus, the tendency and magnitude both showed that the derived haze datasets not only agreed with the meteorological standard but also satisfactorily represented the concentration of the atmospheric composition.
Monthly atmospheric data such as wind, geopotential height, temperature and sea level pressure (SLP) are derived from the National Centers for Environmental Prediction/National Center for Atmospheric Research global reanalysis dataset with a horizontal resolution of 2.5°×2.5° collected from 1979 to 2015 (Kalnay et al. 1996). The monthly mean planetary boundary 100 layer height (PBLH, 1 o ×1 o ) was derived from the ERA-Interim dataset (Dee et al. 2011). The monthly mean Extended Reconstructed SST (ERSST) datasets with a horizontal resolution of 2 o ×2 o collected from 1979 to 2015 were obtained from the National Oceanic and Atmospheric Administration (NOAA) (Smith et al. 2008). The ASI extent was calculated using the ASI concentration data from Hadley Center's HadISST1 with a resolution of 1 o ×1 o collected from 1979 to 2015 (Rayner et al. 2003). The EA/WR and WP indices were computed by the NOAA climate prediction center according to the Rotated 105 Principal Component Analysis used by Barnston et al. (1987). The calculation procedure for the EU index was consistent with that of Wang et al. (2015b). where H500 represents the geopotential height at 500 hPa and overbars denote the area average.
To verify the covariability between the air and external forcings, i.e., SST, ASI and TS, SVD and correlation analyses were applied after linear trends were removed. To some extent, the energy consumption varied continuously and linearly in eastern WHD NCP exhibited rapid increase since winter 2010 with a large trend of 7.36 d/yr. NCP is a haze-prone area in which WHD is distributed nonuniformly (Figure 3a). Two regions exhibited greater WHD: the plain to the east of Taihang Mountains and to the south of Yan Mountains (PETSY) and the south of Shandong Province. As shown in Figure 1, there were four rural stations, three of which were located near the Yan Mountains and were corresponding to less WHD. Another rural site was near the boundaries of Shandong and Henan (BSH) and also resulted in less WHD. Figure 3b shows the WHD anomalies in 2014 with respect to 1979-2012. In addition to a few sites, a larger number of WHD occurred, especially on the BSH (rural area) and the northeast of Hebei. It is notable that WHDs in these two regions show significant increases, filling up the climatic WHD valley as shown in Figure 3a. As a result, the haze-prone area joined together, indicating that the haze 125 pollution was more serious in this region. Actually, the fast increase of WHD in rural area was an obvious reflection of the severe haze disaster in recent years. At the same time, a larger number of WHD occurred on PETSY and the south of Shandong Province. Recently, the haze pollution has become increasingly serious, as shown by the number of haze days and its coverage. The percentages of sites with greater than 30 and 60 hazy days were 71.8% and 51.3%, respectively, in 2014.
As shown in Figure S1, the mean PM 2.5 mass concentrations in Beijing were above 100μg/m 3 on hazy days, which indicated 130 serious pollution in 2014. Although data continuity was influenced by the switch of the observation method in 2013, the observation that WHD NCP in 2014 was greater than before showed robustness. Thus, it was reasonable to treat the year 2014 as a typical case for haze pollution over this region. As shown in Figure 2, NCP experienced the least WHD in 2010, which could be analyzed as an inverse case of the extreme haze phenomena.
In the lower and middle layers of the troposphere, the correlation fields between WHD NCP and H500 (UV850) represented 135 obvious EA/WR and WP patterns (Figure 4a-b). The EA/WR pattern originated from the north-central Atlantic and propagated through Europe, the north of Caspian Sea and North China (Barnston et al. 1987;Liu et al. 2014). The WP pattern showed two activity centers, i.e., the broad area of Southeast Asia and the northwest Pacific and the Kamchatka Peninsula (Barnston et al. 1987). In addition to EA/WR, the EU pattern was another continental Rossby wave train over Eurasia, which significantly impacted the climate of East Asia . Although the EU pattern was unclear in Figure 4a-b, it can be recognized distinctively from the anomalous circulations in 2014 ( Figure 4c). The correlation coefficients between WHD NCP and these three pattern indices were calculated (Table 1). After detrending, WHD NCP showed significant positive correlation with both the EA/WR and WP patterns, indicating the remote impact on WHD NCP from land and sea. The correlation between EU and WHD NCP was initially significantly positive but became insignificant after detrending. Considering that EU could be recognized from low, middle and high layers in winter 2014, the EU pattern was 145 still treated as a possible circulation correlated with WHD NCP . These teleconnection patterns might contribute to WHD NCP by impacting the pivotal and local anti-cyclone anomalies (i.e., the local climate conditions) over NCP (Figure 4a Influenced by the circulations with weaker cold air, horizontal diffusion of the atmospheric particulates was impeded. Near NCP, the anomalous anti-cyclone also existed that illustrated weaker vertical motion. The main physical process was that the climate teleconnection patterns altered the local climate conditions, and then influenced the diffusivity of the local 155 atmosphere. That is, when the positive pattern of EA/WR, WP and EU occurred together or partly, the anomalous anti-cyclone over NCP and Japan Sea was enhanced from surface to the middle troposphere, thus, the convection or vertical motion was confined. The southerly anomalies on the left side of this anti-cyclone weakened the cold air and wind speed, but brought about humid flow. Under such local climate conditions, the vertical and horizontal diffusion of atmospheric particulates were both restricted and then the pollutant gathered in a narrow space that resulted in the occurrence of haze.

160
In winter 2014, many extreme climatic events occurred, such as severe drought and high temperatures in NCP ). Corresponding external forcings should be observed as the background of these extreme synoptic and climatic events, which might persistently impacted the winter atmospheric circulations and led to an irregular distribution of teleconnection patterns in winter. It should be noted that EA/WR pattern in winter 2014 was distributed slightly westwards and broadly.
Nevertheless, the three eastern centers of EA/WR, WP and EU patterns could be recognized ( Figure 4c). The linkage 8 anti-cyclone of these three teleconnection patterns was enhanced and modulated the local climate conditions. The NCP area was influenced by the anomalous high resulting in lower PBLH ( Figure 6). The southerly at the high latitudes deadened the cold air from its main source, so the atmospheric matters gathered easily. Near the surface, the negative SLP anomalies surface wind over NCP (Figure 5f), and the higher PBLH ( Figure 6). The atmospheric diffusivity was heightened by the stronger cold air and vertical movement. This observation further supports the speculation that the anomalous EA/WR and WP patterns contributed to significant changes in WHD NCP by altering the local climate conditions. The anomalies of the EU pattern in winter 2010 were not as significant as those in 2014. As shown by Table 1, the relationship between WHD NCP and 180 the EU index weakened after detrending, illustrating that the correlation was much weaker than the other two patterns.

Possible mechanisms for the winter haze in 2014
In the above discussion, we addressed the associated circulations that were characterized by EA/WR, WP and EU patterns and that contributed to the extreme haze pollution in 2014. Wang et al. (2015a) found that the pre-autumn ASI could significantly impact the WHD in the east of China. As another efficient external forcing, the negative SWP-SST anomalies 185 (SSTA) markedly intensified WHD NCP (Yin et al. 2016a). Thus, the question arises as to whether these factors could have caused the extreme haze pollution that was measured in 2014.
Since 1979, the pre-autumn ASI declined substantially, which might impact the winter climate of East Asia (Liu et al. 2012).
Furthermore, this decreasing trend of the ASI intensified from 2006 . The relationship between the pre-autumn ASI and the circulations was analyzed using the SVD method, explaining 51.6% and 17.9% of the variance by 190 the first (SVD1) and second (SVD2) mode, respectively. The correlation coefficient of the first temporal series was 0.76, which was significantly above the 99% confidence level. The excited anti-cyclonic (A) or cyclonic (C) activity centers of  , which can be recognized as the positive EU phase. When BKLSI was above normal and ESSI was below normal or the surface of mid-latitude Eurasia was warmer, these two centers could be stimulated or enhanced, and represented favorable circulations for haze occurrence. Prior to this study, Yin et al. (2016) found that the pre-autumn SSTA of the Pacific excited the anti-cyclone anomalies over North China and resulted in more WHD NCP . The

215
SST in October and November (ON) and winter H500 were decomposed using the SVD technique to reveal the main relationships. The explained variance and the correlation coefficient of the first component were 60.2% and 0.73, respectively. An EU-like pattern occurred in the first mode of H500, i.e., cyclonic anomalies over the polar area and anti-cyclonic anomalies over Mongolia and North China (Figure 9a). The associated SST was cooler over the northwest Pacific, involving Kuroshio and its extension, and warmer over the central-east Pacific and Alaska Gulf (Figure 9b). Another 220 atmospheric response was the weaker East Asia trough, indicating weaker EAWM and cold air. Figure 9c shows the ON SSTA of the Pacific in 2014, which appeared to be similar to the SST SVD1 distribution and stimulated haze-prone responses. In 2010, the ON SSTA in the Pacific did not show a well-organized opposite pattern ( Figure S4), but a cooler SST was observed on the central-east Pacific and Alaska Gulf, which were of benefit to haze occurrence.
Both EA/WR and EU were continental teleconnection patterns that propagated over Eurasia. In contrast, the WP pattern was 225 located over the junction of the marine and mainland areas, and its positive phases were composed of a broad positive center over the northwest Pacific and Southeast Asia and a negative center over Kamchatka Peninsula (Barnston et al. 1987). From Table 1, the stronger WP positive phase could have contributed to more WHD NCP and could also have been partially responsible for the severe haze event in 2014. To identify the causes of such circulations, we performed an SVD analysis between winter H500 and pre-autumn TS, with explained variance of 39.8% and 20.7%, respectively, for SVD1 and SVD2.

230
The results of SVD2 showed significant anomalies in the two centers of the positive WP pattern and anomalous TS distributed from the southwest to the northeast in Eurasia (Figure 10). The correlation coefficient of the SVD2 time series was 0.62, exceeding the 99% confidence level. Similar to the SVD2 representation, two -southwest to northeast‖ anomalous TS distributions also occurred in autumn 2014, i.e., negative anomalies from the Caspian Sea to Baikal Lake and a positive anomalous belt from Southwest China to Northeast China (Figure 10c). Referring to SVD2, these -southwest to northeast‖ the WP pattern and impacted the local circulations over the NCP area. In autumn 2010, the positive belt of TS from the Caspian Sea to Baikal Lake was significant, and the land surface of the Tibet Plateau was warmer than normal ( Figure S5).
These two anomalies were dramatically inverted with the SVD2 results and could have stimulated a negative WP pattern. As argued by two opposed cases, the mechanism for how the EA/WR and WP patterns and associated external factors impacted 240 the WHD NCP were confirmed.

Conclusions and discussions
Except for a few sites, haze pollution over NCP in

290
Documented by these three case studies, the influences of the highest concentrations of PM 2.5 were the fundamental cause, and the associated with atmospheric anomalies and external forcings played key roles in the severe haze pollution. In this study, we focused on the roles of climate conditions and did not discuss the impact of human activities deeply that should be researched in the future work. To separate the contributions quantitatively by numerical models or advanced statistical approaches would be a meaningful task that was helpful to the interpretation of mechanism and the seasonal prediction (Yin possible that not all of the above factors might be found in a specific case study, i.e., a few of these factors played essential and led to the characteristics of that case. A brief summary of the impacts of these factors on WHD NCP is offered in Table 2 Table 2. Summary of the various influnce factors for WHD NCP . The "+++" indicates "more important"; "++" indicates "important", "+" indicates "less important" , and blank indicates "not important".