Air Stagnations for China ( 1985 – 2014 ) : Climatological Mean Features and Trends

Air stagnation is an important meteorological measurement for unfavourable air pollution conditions, but little is known about it in China. We conducted a comprehensive investigation of air stagnation in China, based on sounding and surface observations of 81 stations, from January 1985 to December 2014. The stagnation criteria were revised to be topographically dependent for the great physical diversity in this country. It is found that the 10 annual mean air stagnation occurrences are closely related to general topography and climate features. Two basins in the northwest and southwest of China—Tarim and Sichuan Basins—exhibit the most frequent stagnation occurrence (50% days per year), whereas two plateaus (Tibet-Qinghai and Inner Mongolia Plateau) and the east coastal areas experience the least (20% days per year). Over the whole country, air stagnations achieve maxima in summer and minima in winter, except for Urumqi, a major city in the northwest of China, where stagnations keep a 15 rather constant value yearly around with a minimum in spring. There is a nationwide positive trend in stagnation occurrence during 1985–2014, with the strongest increasing centres over Shandong Peninsula in eastern China and the south of Shaanxi in central China. Dependence degrees of air stagnations on three components (upperand lower-air winds, precipitation-free days) are examined. It shows that the behaviours of upper-air wind speeds are main drivers of the spatial distribution and trend of air stagnations, near-surface winds the next, and dry days 20 contribute the least.


Introduction
Air quality is strongly dependent on meteorological state, which controls the transport and dispersion of air pollutants within the lower atmosphere.A meteorological state with lingering anticyclones and persistent calm winds leading to poor ventilation and no precipitation to wash out pollutants is defined as an air stagnation event (Wang and Angell, 1999).It has been observed that stagnation events are usually related to air pollution episodes (Jacob et al., 1993;Wang and Angell, 1999;Mickley et al., 2004;Wu et al., 2008;Fiore et al., 2012).For example, ozone measurements from rural sites in the eastern and western United States in 1978 and1979 indicated that the majority of ozone episodes occurred during stagnant atmospheric conditions (Logan, 1989).The stagnation of air masses led to an enhancement of surface ozone and CO mixing ratios over western and central Europe (Ordónˆez et al., 2010;Leibensperger et al., 2008) and a 2.6 µg m −3 increase in fine particulate matter in the United States (Tai et al., 2010).The sensitivity of air quality to stagnation has been investigated by perturbing meteorological variables in regional chemical transport models (Liao et al., 2006).Jacob and Winner (2009) collected and compared results from different perturbation studies on the effects of weather conditions on ozone and particulate matter concentrations; they concluded that air stagnation demonstrates a robust positive correlation.However, the actual air pollution level is affected not only by meteorological conditions, but also other complex factors, such as emission sources and chemical reactions (Cao et al., 2007;Guo et al., 2009;He et al., 2001;Yang et al., 2016).With other factors unknown, the air stagnation index may show a poor correlation with the actual air pollution data in certain situations.The strength of the air stagnation index Q.Huang et al.: Air stagnation in China (1985-2014) is that it provides an independent view of the meteorological background relevant to air pollution without interference from the complexity of other factors, such as the variation in source emissions (Horton et al., 2014).
Air stagnation is identified by thresholds of daily upperand lower-air winds and precipitation (Wang and Angell, 1999).The "upper-air winds" refer to winds at about 5 km above the ground.From a meteorological perspective, this level is important because of its connection to near-surface synoptic systems.It is found that the movement of surface cyclones tends to travel in the direction of the upper flow at roughly one-quarter to one-half speed (Frederick et al., 2012).These kinds of near-surface synoptic systems are essential to air pollution (Jacob and Winner, 2009;Cai et al., 2017).On the other hand, near-surface winds and precipitation determine the dilution and washout of air pollutants, and both are also relevant to practical air quality.Therefore, air stagnation is a relatively simple but conceptually robust metric to air pollution.
In previous work by Korshover (1967) and Korshover and Angell (1982) for the United States, air stagnation events were defined using daily weather maps from the US National Weather Service.They are periods in which (i) the surface geostrophic wind is less than 8 m s −1 , generally corresponding to a 10 m wind speed less than 3.2 m s −1 (Wang and Angell, 1999), and (ii) the wind speed at 500 hPa is less than 13 m s −1 with (iii) no precipitation.Wang and Angell (1999) followed this metric but replaced the dataset with a reanalysis archive (2.5 • × 2.5 • ) from the US National Centers for Environment Prediction (NCEP) National Center for Atmospheric Research (NCAR).With this dataset, they studied the climatology of air stagnation in the United States from 1948 to 1998 and found that air stagnation events happen most frequently in the southern states during an extended summer season from May to October.Based on the work of Wang and Angell (1999), the National Climatic Data Center (NCDC) monitors air stagnation days in the United States with finer-gridded reanalysis data (0.25 • × 0.25 • ) and provides maps on air stagnation distribution every month (http: //www.ncdc.noaa.gov/societal-impacts/air-stagnation/).Following the NCDC metric, Leung and Gustafson (2005) examined the potential effects of climate change on US air quality by analyzing the simulated changes in stagnation events for 2045-2055.Horton et al. (2012Horton et al. ( , 2014) ) furthered this work and used a multi-model ensemble to project future air stagnation occurrence on a global scale.They found that global warming could be expected to result in increasing stagnation frequency over the eastern United States, Mediterranean Europe, and eastern China.
China has experienced rapid economic growth and industrialization in recent decades and become the second largest energy consumer in the world (Chan and Yao, 2008).Tremendous energy consumption results in heavy air pollution.Research on air quality is indispensable, and studies of the relevant meteorological states are essential.To the authors' knowledge, studies on air stagnation have thus far generally focused on the United States or on a global scale.Our primary purpose in the current study is to investigate the climatological mean features and trends in air stagnation in China based on 30-year (1985-2014) observations from stations across the country.
2 Data and methods

Data
Long-term  datasets of daily-mean surface wind speeds (observed at 10 m above the surface) and daily precipitation data were obtained from the China Meteorological Administration (CMA).These data are available from the CMA website (http://data.cma.cn/data/detail/dataCode/SURF_CLI_CHN_MUL_DAY_CES_V3.0.html).Upper-air wind speeds were obtained from the University of Wyoming soundings database (http://weather.uwyo.edu/upperair/sounding.html).This database provides twice daily (00:00 and 12:00 UTC) atmospheric soundings from stations that participate in global data exchange.Daily averages of upper-air wind speeds at mandatory levels of 500, 400, and 300 hPa were used here.
We obtained datasets for all the radiosonde stations across China (95 stations) and two stations (Blagoveshchensk and Vladivostok) outside the country but near the border.Among them, 66 stations have corresponding surface datasets from the CMA (see Appendix A).For each of the other 31 radiosonde stations, we considered the average for the surface stations within 150 km as a substitute.This gave us an additional 15 stations (see Appendix B).Air stagnation data from these 81 stations are analyzed in this study.Figure 1 displays the distribution of the stations.There are relatively fewer stations on the Qinghai-Tibet Plateau, particularly in western Tibet.Despite this, the 81 stations cover all of contiguous China well.

Quality control
Surface datasets have been quality controlled by the CMA (http://data.cma.cn/data/detail/dataCode/SURF_CLI_CHN_MUL_DAY_CES_V3.0.html).Upper-air wind data were eliminated from this analysis if where U i denotes the ith upper-air wind speed at a given mandatory pressure level for a certain station, i ranges from 3 to n − 2, and n is the total number of the data sample.U and σ are calculated as follows: A subjective quality control procedure was also applied.A time series plot of upper-air wind speeds at each station was drawn to screen temporally inhomogeneous data.Under these two quality control procedures, no upper-air wind data have been considered abnormal and therefore removed.
We consider the data used in this study reliable.

Data completeness
A general survey shows that datasets from 73 stations are available from January 1985 to December 2014, while the other 8 stations (Wenjiang, Jinghe, Chongqing, Shanghai, Vladivostok, Yuzhong, Zhangqiu, Qingyuan) cover less than 30 years.The shortest duration is 7 years and 3 months (October 2007-December 2014) at Jinghe station.The percentage of valid data (data for upper-and lower-air winds and precipitation are all valid) for each station is summarized in Appendices A and B. It is shown that for all stations except Blagoveshchensk, more than 95 % of the data are valid.
Overall, the datasets are sufficient to conduct climatological research on air stagnation over China.

Methods
We adjust the NCDC air stagnation index and a given day is considered stagnant when total daily precipitation is < 1 mm (i.e., a dry day), daily-mean surface wind speed is < 3.2 m s −1 , and upper-air wind speed is < 13 m s −1 .In previous studies, the upper-air wind is defined as the wind at 500 hPa.For China, however, this criterion is not appropriate because of the great physical diversity.The Qinghai-Tibet Plateau has an average height of over 4000 m, and wind speeds at 500 hPa are not representative of the upperair winds above the ground.Therefore, we refined the criteria to be topographically dependent, and the mandatory level to provide upper-air wind is chosen according to the station's elevation (Table 1).
With the modified criteria, air stagnation days are identified by checking the meteorological conditions of every day at each station.Furthermore, if there are 4 or more consecutive days of air stagnation conditions at a given station, those days are considered as one air stagnation case (Wang and Angell, 1999).The results for stagnation days and cases were interpolated with cubic splines to a 2 • × 2 • grid to show the spatial distribution over continental China.

Annual occurrence
Annual mean air stagnation days are distributed with substantial regional heterogeneity (Fig. 2a).They are most prevalent over basins in the northwest and southwest of China (i.e., Xinjiang and Sichuan provinces) where stagnant atmospheric conditions account for 50 % of days per year on average.They are less prevalent (about 33 % days per year) over the northernmost and southernmost parts of the country.The remaining regions in China experience even fewer stagnant days, especially the Qinghai-Tibet Plateau, the Inner Mongolian plateau, and the eastern coastal areas where stagnant conditions occur the least (less than 20 % of days per year).The distribution of stagnation cases agrees well with that of stagnation days (Fig. 2b).The strongest stagnant centers in the Xinjiang and Sichuan basins exhibit more than 16 cases per year, while the weakest centers on the Qinghai-Tibet Plateau, the Inner Mongolian plateau, and the eastern coastal areas only experience fewer than 2 cases per year.Air stagnation cases usually persist for about 5 days in a majority of these areas (Fig. 2c).Those over the basins of Xinjiang and southern China last longer (6 days).The longest duration of air stagnation conditions occurs in the south of Guangxi, lasting more than 7 days.

Seasonal occurrence
Generally, most stagnant air stagnation conditions happen during the summer season, while only a few occur during Table 1.Station elevations and the corresponding upper-air wind speed criteria.

Station elevation Topographically dependent upper-air (m)
wind speed criterion 0-1000 Wind speed at 500 hPa < 13 m s −1 1000-3000 Wind speed at 400 hPa < 13 m s −1 3000-4000 Wind speed at 300 hPa < 13 m s −1 winter.Stagnation days in autumn are slightly more frequent than in spring (Fig. 3).A similar feature was also found in the earlier work of Wang and Angell (1999) for the United States.The seasonal variation in stagnation is attributed to a seasonal shift in pressure patterns and general circulation.A much weaker pressure gradient in summer is a well-known seasonal feature in the midlatitudes (Frederick et al., 2012).This feature is very evident in the upper layer of the atmosphere in China (Ding et al., 2013 and their Fig. 1.1).However, at sea surface level, the case in eastern Asia and China is complicated by the subtropical high in the east and the continental low and in the west.As a result, the Asian summer monsoon prevails in eastern China.Except for this, the sea level pressure gradient in summer is still much weaker than in winter (Ding et al., 2013 and their Fig. 1.1).A weaker wind in both the upper and surface layer accompanies the weaker pressure gradient and results in more air stagnation occurrences in China and North America (Wang and Angell, 1999).We choose four stations (Harbin, Urumchi, Beijing, and Chongqing; shown in Fig. 1 as triangles) and the average for China (all stations in this study) to demonstrate the seasonal variation in air stagnation days and cases.Figure 4 shows that for all but Urumchi, stagnation days and cases begin to increase in March or May, grow dramatically and achieve maxima in July or August, and then fall sharply and reach minima in December or January.However, monthly stagnation days and cases for Urumchi show much less variation in a year with their minima in April.This may be attributed to the unique local climate there.Xinjiang basin is isolated at the center of the continent, far away from the coast and blocked by the huge Qinghai-Tibet Plateau in the south.Therefore, the eastern Asian monsoon, particularly the summer monsoon that prevails in eastern China, has little influence on Xinjiang.As a result, the seasonal feature of the stagnation in Urumchi is different from that in eastern China.By comparing these stations, we find that stagnation over Chongqing, Beijing, and Urumchi is higher than the average level of the entire country.Chongqing station has the largest variation in stagnation, followed by Beijing.Urumchi maintains a relatively high stagnation frequency throughout the year.

Trends in stagnation
The majority of China exhibits positive trends with about 10-20 stagnation days and 1-3 stagnation cases per decade (Fig. 5a, b).The strongest centers are located in Shandong Province and southern Shaanxi, with rising rates of more than 20 days and 3 cases per decade.Only four areas exhibit a weak decrease: the extreme north of China, regions located in southern Gansu and northern Sichuan, and the westernmost and southernmost parts of China.The negative trend in stagnation varies from 0 to 10 days and 0 to 1 case per decade over the first three regions and 30 days and 5 cases per decade over the last region.We have assessed the statistical significance of the above results, and 52 % of stations passed the 0.05 significance test.Not only stagnation frequencies show an increase over large areas, but the duration of stagnation cases also exhibits a nationwide extension of about 0.3 day decade −1 (Fig. 5c).Only a few scattered regions show a gradually shortened stagnation duration, including the extreme north of China, the Yangtze River Delta, and the westernmost and southernmost regions.Stagnation trends for four stations (Harbin, Urumchi, Beijing, and Chongqing) are also specifically discussed along with the average results for the entire country.Figure 6 shows that all four stations exhibit positive trends in air stagnation days, ranging from 3 to 14 days decade −1 , and the nationally averaged increasing rate is about 6 days decade −1 .Moreover, the two stations experiencing more stagnant days (Urumchi and Chongqing) exhibit a slower rising rate of about 5 and 3 days decade −1 , respectively, whereas the other two stations having relatively less stagnation (Harbin and Beijing), showing a faster rate of about 14 and 8 days decade −1 , respectively.

Discussion
Air stagnation is identified based on the thresholds of three components: the lower-and upper-air wind speeds and precipitation-free days.An analysis of each individual component is helpful in understanding the behavior of stagnation.Figure 7a shows that the distribution of the annual mean upper-air wind field is reversely correlated with that of stagnation occurrences.The upper-air wind is relatively weak in northeastern China, the Tarim basin, the Sichuan basin, and the southernmost part of the country; these are exactly the same four regions exhibiting frequent stagnation.In contrast to the pattern of the upper-air wind field, the surface wind field exhibits a strong wind center (> 3 m s −1 ) around northeastern China and a weak one (about 1.5 m s −1 ) in the Sichuan basin (Fig. 7b).The surface wind speed over the remainder of China is around 2 m s −1 .The distribution of dry days (daily total precipitation < 1 mm) is largely related to the latitude.Figure 7c shows that dry days generally occur more often in the north than in the south.Specifically, southern Xinjiang Province experiences a maximum of more than 350 dry days per year, while southern Shaanxi Province shows a minimum of about 200 days per year.
We further analyze the degrees of stagnation dependence on each individual component (Fig. 8).The result implies that 76 % (R 2 ) of the spatial variation in stagnant days can be explained by the distribution of upper-air wind fields.The correlation even reaches as high as 82 % in autumn (Fig. S1 in the Supplement).The surface wind field only accounts for 20 % and the spatial distribution of dry days barely influences the stagnation variation.Figures 7 and 8 show that stagnation occurrences result from the cumulative responses of individual stagnation components, but the distribution of upper-air wind speeds exerts the dominant influence.The same feature was also suggested in the global research by Horton et al. (2012) in the area of China.
Similarly, we examine the relationship between stagnation trends and each component (Fig. 9) and find that the pattern of trends in the upper-air wind field is similar to that of stagnant conditions.Decreases in the upper-air wind field substantially outnumber increases throughout the country (Fig. 9a), and the regions showing rapidly decreasing winds coincide with those exhibiting robust growing stagnation in Fig. 7. Trends in surface wind field and dry days may show a slightly different pattern from trends in stagnation (Fig. 9b,  c), but they still contribute more or less for some regions.For areas with increasing stagnation, like northeastern and southern Shaanxi Province, dry days show a substantial positive trend (about 3-7 days decade −1 ) and upper-and lowerair wind speeds show a remarkable negative trend (about 0.3-0.6 m s −1 decade −1 ).For the Shandong region, both upperand lower-air wind speeds exhibit a substantial decrease of more than 0.3 m s −1 decade −1 , although the dry days show a slightly decrease of about 1 day decade −1 .To summarize,  the stagnation trends are contemporaneous effects of two or three components.
The dependence of nationally averaged stagnation trends on each individual component is shown in Fig. 10.It can be seen that the negative trend in the upper-air wind speed accounts for 73 % of the increase in stagnant days.The ratio varies slightly with seasons, and the highest (79 %) occurs in spring (Fig. S2 in the Supplement).Interannual variations in surface wind speed and dry days explain 42 and 32 %, respectively.Still, trends in the upper-air wind are the dominant contributor.
The trends in these three components (upper-air winds, near-surface winds, and daily precipitation) are driven by climate change.The decrease in upper-air winds results from smaller contrasts of the sea level pressure and a weakened Hadley circulation, both as a consequence of global warming (Lau et al., 2006;Lu et al., 2007;Seidel et al., 2008).The near-surface wind decline is attributed to the slowdown in atmospheric general circulation (Guo et al., 2011;Xu et al., 2006;Vautard et al., 2010) and the stabilized atmosphere by light-absorbing aerosols (Li et al., 2016;Peng et al., 2016;Wang et al., 2013).The decreasing number of rainy days, due to suppressed light rainfall but intensified heavy rainfall, is mainly attributed to the accumulation of greenhouse gases and aerosols in China (Gong et al., 2004;Liu et al., 2015;Wang et al., 2011aWang et al., , 2016)).To sum up, climate changes alter atmospheric circulation and the hydrological cycle, which influence the occurrences of air stagnation as the meteorological background of air quality.
Stagnation is a meteorological metric for potential air pollution occurrence.Once there are anthropogenic or natural air pollutants, they are likely to accumulate and result in poor air quality over regions that experience frequent stagnant conditions.In contrast, over regions with infrequent stagnation, air pollutants will quickly be transported far away and diluted.The current results from the prevalent centers of stagnation days and cases are consistent with areas of heavy pollution in China (Mamtimin and Meixner, 2007;Wang et al., 2011b;Chen and Xie, 2012;Liu et al., 2013;Zhang et al., 2014;Li et al., 2015).The spatial distribution of annual mean visibility during 1985-2014 (Fig. S3 in the Supplement) shows that regions in the Sichuan basin, western Xinjiang, and the North China Plain exhibit low visibility.This feature corresponds well to the frequent air stagnation occur-  It should be noted that the air stagnation metric does not take into account emissions or atmospheric chemical reactions, so there may be discrepancies between variations in the stagnation index and the actual air pollutant concentrations in certain situations.For example, the aerosol concentration in China is characterized by a high value in winter and a lower value in summer.This observational fact is clearly related to the seasonal variation in source emissions, since there is more coal consumption in winter for heating, particularly in northern China (Cao et al., 2007;He et al., 2001).To make this kind of meteorological metric applicable for practical air pollution forecasting, Yang et al. (2016) incorporate source emission information into their PLAM index and successively improve the forecasting skill.In this work, we aim to analyze the general features of the meteorological background relevant to air pollution by means of the air stagnation metric without concern for the complexity of source emissions or chemical reactions.

Conclusions
Based on upper and surface wind speeds and daily precipitation data from 81 stations across the country, this paper presented climatological mean values and trends in air stagnation in China from January 1985 to December 2014.The dependence of stagnation on three components (upper-and lower-air winds and dry days) was examined.A topographically dependent version of air stagnation criteria was applied to account for the terrain effect in China.
The annual mean air stagnation occurrence varies spatially, which is in agreement with topography and climate features.Two basins in northwestern and southwestern of China, the Tarim and Sichuan basins, exhibit the most frequent stagnation occurrence (50 % of days per year).Two plateaus (the Qinghai-Tibet and Inner Mongolian plateaus) and the eastern coastal areas experience the least (20 % of days per year).Seasonal variation in air stagnation is also presented.For a general view of the whole country, stagnation happens most frequently in summer and least frequently in winter.For specific stations in Harbin, Beijing, and Chongqing, stagnation varies dramatically by month and achieves maxima in July or August and minima in December or January; stagnation in Urumchi maintains a rather constant value with a minimum in April.
There is a nationwide positive trend in stagnation days, stagnation cases, and case duration during 1985-2014.The strongest increasing centers are located over Shandong Province in eastern China and southern Shaanxi in the middle of the country.Only two regions in the southernmost and westernmost parts of China exhibit a negative trend in both occurrence and duration.
Stagnation occurrence contemporaneously responds to three components: upper-and lower-air winds and precipitation-free days.Among these, the upper-air wind speed plays a dominant role, explaining 76 and 73 % of the spatial distribution and trends in air stagnation, respectively.The lower-air wind exerts a minor influence.These results are corroborated by the global research by Horton et al. (2012).The spatial variation in dry days barely influences stagnation, whereas interannual variability explains 32 % of the stagnation trend.
Air stagnation climatology presents a specific view of the natural background of the atmospheric features responsible for air pollution levels.The results presented in this paper may have significant implications for air pollution research and may be used in atmospheric environmental management or air pollution control.

Figure 1 .
Figure 1.Distribution of the observation stations.The triangles indicate the four stations selected to discuss seasonal variations in air stagnation in Sect.3: 1 is Harbin, 2 is Urumchi, 3 is Beijing, and 4 is Chongqing.The squares indicate the two stations outside of China at Blagoveshchensk (a) and Vladivostok (b).
Figure 2. Annual mean air stagnation days (a) and cases (b) and the mean duration of stagnation cases in days (c) for China (1985-2014).

Figure 4 .
Figure 4. Seasonal cycles of monthly mean air stagnation days and cases for four stations (Harbin, Urumchi, Beijing, and Chongqing) and all of China.The solid line indicates stagnation days, and the dashed line indicates stagnation cases.

Figure 5 .
Figure 5. Trends in stagnation days (a) and cases (b) and the duration of stagnation cases (c) during 1985-2014.

Figure 6 .
Figure 6.Annual mean stagnant days and the corresponding trends at four stations (Harbin, Urumchi, Beijing, and Chongqing) and for the whole country.

Figure 8 .
Figure 8. Dependence of the spatial distribution of stagnation days on three components (upper-air wind speed, surface wind speed, and dry days).Linear regression coefficients between annual mean stagnation days at 81 stations and each corresponding component are shown.

Figure 10 .
Figure10.Same as Fig.8, but for the trends in stagnant days.Linear regression coefficients between nationally averaged stagnant days in the 30-year period and each corresponding component are shown.
* World Meteorological Organization identification number.