Re-evaluating the variation in trend of haze days in the urban areas of Beijing during a recent 36-year period

By using meteorological station data, the inter-annual variability of haze days and its trends are re-evaluated in the urban areas of Beijing during a recent 36-year period. Observations from station 54,511, which is a national reference climatological station in the urban area of Beijing, are not suitable for representing the whole urban area, since it trends oppositely to the surrounding stations. Instead, averaged haze days according to five stations in the urban area of Beijing were selected rep-resentatively, illustrating that haze days have a positive trend during the period 1980 – 2015, and haze occurs more often in autumn and winter than in spring and summer. Notably, the number of haze days has increased more rapidly in summer than in the other three seasons. Severe and persistent haze days exhibit positive trends of 4.1 and 13 days/decade, respectively, during the investigation period, while the corresponding ratios to the total haze days have also increased gradually. The haze in Beijing has also become more severe and drier. Under the weakening East Asian monsoon in winter, there has been a reduction in days of locally strong wind speeds and rain, and an increase in days of weak wind speeds. This has directly contributed to the weakening of the diffusion of pollutants, which would otherwise act to maintain the haze, thus prolonging the duration of haze pollution days in urban areas of Beijing.

By using meteorological station data, the inter-annual variability of haze days and its trends are re-evaluated in the urban areas of Beijing during a recent 36-year period. Observations from station 54,511, which is a national reference climatological station in the urban area of Beijing, are not suitable for representing the whole urban area, since it trends oppositely to the surrounding stations. Instead, averaged haze days according to five stations in the urban area of Beijing were selected representatively, illustrating that haze days have a positive trend during the period 1980-2015, and haze occurs more often in autumn and winter than in spring and summer. Notably, the number of haze days has increased more rapidly in summer than in the other three seasons. Severe and persistent haze days exhibit positive trends of 4.1 and 13 days/decade, respectively, during the investigation period, while the corresponding ratios to the total haze days have also increased gradually. The haze in Beijing has also become more severe and drier. Under the weakening East Asian monsoon in winter, there has been a reduction in days of locally strong wind speeds and rain, and an increase in days of weak wind speeds. This has directly contributed to the weakening of the diffusion of pollutants, which would otherwise act to maintain the haze, thus prolonging the duration of haze pollution days in urban areas of Beijing.

K E Y W O R D S
Beijing, haze events, inter-annual variability, long-term trends, representative station

| INTRODUCTION
The recent increase in haze pollution in China has strongly affected the health of the population, which is mainly the result of the increasing amount of emissions as the population increases and the economy grows rapidly, combined with poor diffusion conditions (Guo et al., 2011Liu et al., 2013;Jiang et al., 2015;Liu et al., 2015;Wang and Chen, 2016;Yang et al., 2018). Air quality has a significant influence on both social and economic development, public health, and climate change (Huang et al., 2014;Li et al., 2016;Gu et al., 2018;Hu et al., 2018;Wang et al., 2019).
Beijing is one of the fastest developing cities in China, and also one of the cities suffering the most from severe air pollution. The PM 2.5 concentration even reached 886 μg/m 3 in January 2013, drawing significant attention from the government and the wider society Wang et al., 2014aWang et al., , 2014b. Research into haze in the Beijing-Tianjin-Hebei region (BTHR) has mainly focused on two aspects: on the one hand, anthropogenic emissions are a major factor, including biofuel burning, traffic, industrial and dust emissions, and secondary pollutants (Zhang et al., 2007;Watson et al., 2008;Liu et al., 2014). On the other hand, previous work has shown that the weather conditions determine the pollutant diffusion, depending on the wind speed, wind direction, relative humidity, assuming pollutants are released steadily (Ding and Liu, 2013;Zhang et al., 2014). In addition, inter-annual variability and the long-term trend of heavy haze pollution are sensitive to the natures of the local and large-scale circulations Ji et al., 2012;Chen and Wang, 2015;Miao et al., 2015aMiao et al., , 2015bMiao et al., , 2017Cai et al., 2017;. Therefore, for the establishment of observationally long-term variations and trends of haze days, it is important that a reference observation station is suitably selected to explore the haze trend and its relationship with climate variability.
In recent years, Hu and Zhou (2009) and Wu et al. (2014) found a negative trend of haze days from 1980 to 2005 based on the data of station 54,511, which is a national reference climatological station in the urban areas of Beijing. However, this differs from the trend in the whole BTHR Ding et al., 2017). Therefore, it is necessary to investigate whether the trend in Beijing truly differs from the whole region, or whether this inconsistency results from artifacts in the data. The primary goals of this paper are to substantiate this controversy and determine the intrinsic nature of the long-term and inter-annual variability of haze in the urban areas of Beijing.

| DATA AND METHOD
Daily data of meteorological stations in Beijing, Tianjin, Langfang, and Tangshan from 1980 to 2015 are analyzed, which measure the relative humidity, visibility, wind speed, precipitation, and other meteorological variables. All data were provided by the Beijing, Hebei and Tianjin Meteorological Information Centers. Here, the East Asian winter monsoon (EAWM) index is defined by the mean geopotential height at 500 hPa in the area of (25 -45 N, 110 -145 E) to describe the East Asian trough (EAT) intensity, which is closely associated with the EAWM and cold activity (Wang and He, 2012;Wang et al., 2015); the larger the value of EAWM index, the weaker the EAWM. The annual energy consumption data in Beijing is obtained from the Bureau of statistics of Beijing (http://www.bjstats.gov.cn/).
There are two common methods to define a haze day (Schichtel et al., 2001;Doyle and Dorling, 2002;Che et al., 2009;Wu et al., 2010). One definition of haze (hereafter refer to daily-average method) is when the daily average visibility and relative humidity are less than 10 km and 90%, respectively, excluding natural events such as precipitation, dust, fog, mist, and gales using the present weather code. The second definition (hereafter refer to 14-hr method) is based on the observation at 1400 local time instead of a daily average, because the midday values are more representative of the regional visibility levels, as early morning radiation fog and the high relative humidity, which may reflect only local conditions, would have mostly dispersed by midday (Lee, 1990;Doyle and Dorling, 2002;Che et al., 2009;Wu et al., 2010). For instance, Zhao et al. (2011a), Zhao et al., 2011b) compared these two methods, and concluded that the latter is better for haze-day identification, especially for large regions and a long temporal analysis.
Therefore, the 14-hr method is used here for the analysis on the variation in trend of haze days in the urban areas of Beijing during a recent 36-year period. Additionally, a severe haze day is defined as a day with a visibility <3 km, and continuous haze days refers to haze lasting more than 2 days according to the observational criteria from the China Meteorological Administration (QX/T 113-2010). A mixed haze-and-fog day is referred to as a day of relative humidity of 80 to 95%, and a visibility of less than 10 km, excluding natural events.

| Selection of representative sites
Of the 20 meteorological stations in the Beijing area ( Figure 1a), one station (54511) serves as the national reference climatological station and is usually also used as the representative of this area. Since 1980s, there have been two relocations at the 54,511 station: Figure 1b shows a negative trend of haze days of 25.9 days/decade during the period from 1980 to 2015 according to data from station 54,511. In contrast, the average number of haze days according to the other 19 stations gives a positive trend of 11.1 days/decade in the same period. However, energy consumption, which is to some degree an indicator of pollutant emissions, has risen exponentially since 1980, and is negatively correlated with the number of haze days observed at station 54,511 (correlation coefficient: r = −0.75, p < 0.01).
The correlation between the haze days in Beijing and other cities nearby is shown in Table 1, which are cities with similar pollutant emissions, and possess the same climatic conditions. At station 54,511, haze days are negatively correlated with those observed at the Tianjin, Langfang, and Tangshan stations. In contrast, the average number of haze days at the other 19 stations in Beijing has a positive correlation with the three aforementioned stations, which means these stations display the same positive trend for the number of haze days. Such trends are consistent with previous studies in northern China, and even eastern China, where the number of haze days indicates the same positive tendency (Song et al., 2013;Zhang et al., 2015;Ding et al., 2017). Thus, the observations at station 54,511 are not suitable to represent the variation in the number of haze days in Beijing.
According to the definition of a haze day, the relative humidity observed at station 54,511 is slightly lower than that averaged from the other 19 stations in Beijing, with both varying consistently (Figure 2a). Nonetheless, the visibility from the above two datasets is quite different, showing the opposite trends before 2005 ( Figure 2a). As station 54,511 has been moved two times since 1980, which probably caused the inhomogeneous series of visibility and relative humidity at different locations throughout the studied period (Li and Yan, 2009;Yan et al., 2010;, it is invalid to use observations at station 54,511 to represent the variations in haze days in the Beijing area.
Referring to empirical orthogonal functional analysis (Ren et al., 2008), and excluding the stations moved two times or above since 1980, or those located over 300 m above the sea level, we finally chose the stations FS, SY, HD, SJS, and DX as the reference urban stations from other 19 stations (see Figure 1a), and then calculate their average haze days as be representative value of the Beijing urban area. In addition, the SDZ station with no relocation history is selected as a rural station since it is far from the sources of urban pollutants and is the only suitable station for a background reference in northern China. The correlations of haze days between the average of the five stations and those stations at Tianjin, Langfang, and Tangshan are 0.79, 0.79, 0.85, respectively, and the average number of haze days of the five stations is highly correlated (0.78) to the energy consumption of Beijing. The SDZ rural station has a greater correlation to the stations at the aforementioned three cities (0.63, 0.62, and 0.79, respectively). Therefore, the analysis Note that SJS station moved once on January 1, 1998, but the relocated distance was less than 1 km, and FS station was also relocated once on January 1, 2006, and the relocated distance was less than 2 km. There were no significant changes in underlying surfaces for both these station before and after their relocations, and also no discontinuity for observation series were tested by using the homogeneity method  Additionally, the change rate of urban haze days is 15.8 days/decade, which is significantly more than the rural rate (11.9 days/decade). Such differences are probably the result of the increased amount of pollutants produced in urban areas than in rural areas. In particular, the decrease in the urban-rural difference during the 2008 Olympic Games emission-reduction program also shows that local emissionreduction efforts can still prove beneficial. Notably, comparing with the daily-average method, here as the 14-hr method is used to reconstruct haze days in Beijing area, it will underestimate the number of haze days, but it can still reflect the long-term trends of haze days (Zhao et al., 2011a(Zhao et al., , 2011b.

| Urban-rural difference in haze days
3.3 | Inter-annual variation in seasonal haze days  May), summer (June-August), autumn (September-November), and winter (December-February), the average numbers of haze days in Beijing are 9.7, 11, 16.5, and 15.9 days, respectively, accounting for 17.8, 19.6, 31.7, and 30.9% of the haze days over an entire year, implying haze occurs more frequently in autumn and winter than in spring and summer. The differences in seasonal haze days can be explained as follows: because Autumn and winter are traditional heating and coal-burning seasons in northern China, the haze days in autumn and winter are relatively higher because of the concentrated population and the consequent increased energy consumption and unfavorable meteorological conditions (Cai et al., 2017;Pei et al., 2018;Yang et al., 2018;. While the lower haze days in spring are closely related to the large average wind speed and more windy days during this period, and those in summer are mainly corresponding to more wet deposition of precipitation (Miao et al., 2015b;. Over the investigated period, there is a positively increasing trend in each season, while the trends in the percentages of haze days in each of the four seasons are distinctive. There is no obvious trend in percentages in spring ( Figure 3a); in summer, there is a positive trend with a rate of 3% per decade ( Figure 3b); in autumn and winter, the percentages of haze days appear to decrease slightly (Figure 3c,d). More efforts may be taken into the research of such trends, especially on the trends in summer, which is worth exploring in future work.

| Long-term trends in severe and persistent haze days
Both severe haze days in Beijing and its percentage exhibit positive trends with the rates of 4.1 and 4.3%/decade from 1980 to 2015 (Figure 4a). Recently, continuous haze days occur more frequently, and have gained significant attention among citizens. For instance, haze in the Beijing-Tianjin-Hebei region continued for about 1 week in January 2013, representing one of the severest air-pollution events in the recent 5 years. Figure 4b shows the persistent haze days in the Beijing area, whose percentages have increased gradually with some fluctuations from 1980 to 2015. The number of persistent haze-day amounts to over 70 in 2014 and 2015, which far exceeds the average number of haze days (33) during the investigated period. Moreover, persistent haze days account for more than 70% of the total haze days in 2014 and 2015. In both the urban and rural areas in and around Beijing, the number of persistent haze days has increased by 13 and 8.3 days/decade, respectively; the corresponding percentages of total haze days have also increased by 7 and 8.8% per decade, respectively. In general, we deduce that haze in Beijing will probably become more serious and longer duration during1980-2015, the reason of which will be discussed in Section 3.5.

| Impacts of meteorological conditions on haze days in Beijing
In addition to pollutant emissions Guo et al., 2011Guo et al., , 2016, meteorological conditions also play an important role in air quality, especially when the emission rate is steady. On one hand, many previous studies have reported that variations in large-scale atmospheric circulation can significantly influence the air pollution in China via directly affecting the pressure distribution, precipitation, humidity, and flow field near the surface, which in turn affect the advection of air pollutants, such as that associated with the weakened EAWM and its inter-annual variation (e.g., Chen and Wang, 2015;Li et al., 2016;Zhang et al., 2016;Cai et al., 2017;Pei et al., 2018;Yang et al., 2018). Figure 5a shows the inter-annual variations and trends of haze days in Beijing and of the EAT index for 1980-2015, when the EAT index has largely trended negatively, particularly in the 1980s and 1990s, when the decline was significant. The correlation coefficient between the haze days in Beijing's urban areas and the EAT index is −0.6, which is a significant negative correlation (p < 0.01), suggesting that the EAWM is weakening with a shallow East Asian trough, which results in less cold activities, a sinking air motion in the middle-lower troposphere and more stagnantly synoptic circulations in planetary boundary layer, so it is not conducive to the vertical and horizontal dispersion of pollutants in Beijing (Wang and He, 2012;Wang et al., 2015;Zhang et al., 2016;Cai et al., 2017;Guo et al., 2017;Pei et al., 2018;Yang et al., 2018). However, the correlation coefficient between haze days at Station 54,511 and the EAT index is 0.19, showing a positive correlation, which does not correspond with the actual situation. Therefore, it is unreasonable to apply the uncorrected data at 54511 to characterize the long-term variation in haze days in urban areas of Beijing.
Local meteorological parameters, such as the relative humidity, visibility, wind speed, and precipitation, modulate the formation of polluted haze episodes (Wang and He, 2012;Ding and Liu, 2013;Zheng et al., 2015;Wang and Chen, 2016;Liang et al., 2017;Miao et al., 2017). Figure 5b shows that the visibility in Beijing has been declining, with a decrease of 1.1 km/decade in terms of the daily-averaged visibility. In the recent decade (2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015), the average visibility was ≈6.8 km, which is significantly lower than that in the 1980s (≈9.6 km), indicating that there has been a continuous worsening of haze pollution recently. The relative humidity has also trended downward on the whole, dropping during haze days by 1.4%/decade, indicating the development of dry-haze days (Figure 5c). The mixed haze and fog days have significantly reduced by 5.5 days/decade, indicating the reduced conversion of haze to fog (Figure 5c). Moreover, the wind speed has decreased gradually at a rate of 0.15 m/s per decade. Days of strong wind speed (daily maximum wind speed ≥8 m/s) have also reduced at a rate of ≈6.8 days/decade, while weak windspeed days (daily maximum wind speed ≤4 m/s) have increased at a rate of ≈24.5 days/decade (Figure 5d). Haze days present a significant negative correlation with strong wind-speed days (correlation coefficient: r = −0.52, p < 0.01), but a positive correlation with weak wind-speed days (r = 0.3, p < 0.1). In addition, there is a significant negative correlation between haze days and rain days (r = −0.44, p < 0.01, figure not shown). Hence, haze days in Beijing are becoming more severe and drier. In recent years, generally, the reduction in days of local strong wind speed and of rain, as well as the increase in days of weak wind speed, have directly contributed to the weakening of the diffusion capacity of pollutants, which are conducive to haze maintenance, prolonging the duration of haze-pollution days in Beijing.

| CONCLUSIONS AND DISCUSSIONS
Haze pollution has occurred more frequently in recent years in the Beijing area, which has thus attracted greater attention from the government and wider society. Observations from station 54,511 are generally used to study this phenomenon in the Beijing area. This study has confirmed that the time series of haze days from station 54,511 are not suitable for the investigation of the inter-annual variability and the longterm trends in haze pollution in Beijing. Instead, time series First, comparing haze days derived from station 54,511 with other stations in the vicinity verifies that the station 54,511 is not suitable to represent the haze in the urban area of Beijing, as the haze days derived from its record show the opposite trend to that from other stations.
Secondly, haze days in the urban area have occurred more frequently than those in the rural areas, increasing by 15.8 days/decade, compared with the 11.9 days/decade in rural areas. Haze days have a positive trend during the period of 1980 to 2015, with haze occurring more often in autumn and winter than in spring and summer. In summer, haze days have been increasing faster than the other three seasons. Moreover, severe and persistent haze days respectively have been increasing by 4.1 and 13 days/decade, corresponding to a positive 4.3 and 7%/decade based on the total number of haze days.
Finally, the correlation coefficient between the number haze days in Beijing's urban areas and the EAT index is −0.6. Since the EAWM has been weakening with a shallow East Asian trough, which causes lesser cold activities, a sinking air motion in the middle-lower troposphere and more stagnantly synoptic circulations in planetary boundary layer, and these conditions are not conducive to the dispersion of pollutants. However, the correlation coefficient between haze days at station 54,511 and the EAT index is 0.19, which is a positive correlation, and which contradicts the known situation. For this reason, it is unreasonable to apply the uncorrected data at station 54,511 to characterize the variation in haze days in Beijing. In addition, hazes in Beijing are becoming more severe and drier. Within the investigated period, the reduction in days of locally strong wind speeds and rain, and the increase in days of weak wind speeds have directly contributed to the weakening of the diffusion of pollutants, prolonging the duration of haze events in Beijing.
However, there are many factors that influence the continuity, representativeness and accuracy of the observation results, such as multiple station relocations, as well as changes in observational methods Yan et al., 2010). As a result, station 54,511 does not objectively reflect the variation in haze days in the urban areas of Beijing. In contrast, the averaged results from the five selected urban stations in the present investigation are more realistic as representative sites for studying haze days in the urban areas of Beijing during the investigated period, while also FIGURE 5 Inter-annual variations and trends of (a) EAT index with haze days at the 54,511 and selected five urban stations, (b) annual average visibility and average visibility in haze days, (c) total mixed days of fog and haze, annual average relative humidity, and average relative humidity in haze day, (d) annual average wind speed, and average wind speed in haze day, total days of strong (daily maximum wind speed ≥8 m/s) and low wind speed (daily maximum wind speed ≤4 m/s) day demonstrating some new phenomena. For instance, the growth rate of haze days in the summer is most noticeable among the four seasons, with severe and persistent haze days showing significant positive trends. Although the government has adopted a series of policies to reduce emissions in recent years, due to the unfavorable meteorological conditions (Ji et al., 2012;Cai et al., 2017;Ding et al., 2017;Guo et al., 2017;Wu et al., 2017;Pei et al., 2018;Yang et al., 2018), the deterioration of the atmospheric conditions will still be a continuing problem in the urban areas of Beijing.