1 Introduction

The eastern part of China, especially the North China Plain (NCP), has frequently suffered large-scale, severe haze-pollution events in recent years due to rapid urbanization, economic expansion, and industrialization (Chan and Yao, 2008; Wang et al., 2016 ; Yang et al., 2016 ; Yin and Wang, 2016). The high concentration of particulate matter and poor visibility have been a serious threat for human (Deng et al., 2014 ; Vander Zee et al., 2016 ) and ecological health (Zhang et al., 2010 ; Chen et al., 2014 ). Many studies on physicochemical properties and the formation mechanism of winter haze pollution on the NCP have been conducted using different methods, such as satellite remote sensing and ground observations (Guo et al., 2014 ; Li et al., 2016 ; Sun et al., 2016 ; Chen et al., 2017 ), numerical models (Fu et al., 2016 ; Long et al., 2016 ; Yang et al., 2016 ), and some statistical analyses (Batterman et al., 2016 ; Zhang J et al., 2016 ; Zhao et al., 2016 ). The hygroscopic growth of high levels of secondary inorganic aerosols during haze episodes and the weakening of winds over the years were found to be dominant factors leading to the decadal increase in winter aerosol concentrations and haze days over eastern China during recent decades (Yang et al., 2016 ; Tie et al., 2017 ). To improve urban air quality and solve traffic congestion, a series of emission-restriction measures has been implemented during several recent important events in Beijing and surrounding regions, e.g., the 2014 Asia–Pacific Economic Cooperation (APEC) summit and the 2015 Grand Military Parade, commonly referred to as "APEC Blue" and "Parade Blue", respectively, and has been shown to be an effective measure of emission reduction (Chen et al., 2015 ; Huang et al., 2015 ; Liu et al., 2016 ; Sun et al., 2016 ; Wang and Dai, 2016). All aerosol species and NO₂ columns were shown to have significant decreases of 40% to 80% during APEC from those measured before APEC (Chen et al., 2015 ; Liu et al., 2016 ; Sun et al., 2016 ), indicating that emission control on a regional scale substantially reduced air-pollutant levels.

Lanzhou, the capital of Gansu Province, is located in the northwest of China and has experienced heavy air-pollution issues since the 1960s, as a result of large-scale emissions from the petrochemical industry and the valley basin, which is conducive to forming stagnant meteorological conditions (Zhang et al., 2000 ). The Xigu District of Lanzhou City is an important petrochemical and other major industrial base and also the first place in China that photochemical smog occurs (Zhang et al., 2000 ; Gao et al., 2004 ). The air-pollution characteristics in urban Lanzhou and its formation mechanisms were studied intensively by previous researchers using methods of statistical analysis and numerical modeling (Zhang and Hu, 1992; Chen et al., 1994 ; Hu and Zhang, 1999; Savijärvi and Jin, 2001; Zhang, 2001; Zhang et al., 2001 ; Zheng et al., 2006 ; Zhang and Li, 2011). Half of the days in winter exceeded the national grade II standard (Chen et al., 2010 ). It was shown that the daytime temperature inversion induced by aerosol-meteorology feedback and the peak-heating effect and calm wind conditions in the lower atmosphere due to valley terrain were two key factors in controlling air-pollution concentration. Although some proposals and measures such as mountain-slope afforestation and changing local terrain were able to break the long-standing boundary-layer temperature inversion, there was little improvement in air quality (Chu et al., 2008a ; Yu et al., 2016 ). However, with implementation of the Air Pollution Prevention Battle (APPB) in 2012 and the Plan for the Control of Air Pollution (PCAP) in 2015 in Lanzhou, the air quality of urban Lanzhou improved (http://news.qq.com/a/20130115/000007.htm) and appeared frequently in the phenomenon of "Lanzhou Blue" (Figure 1). Lanzhou City was awarded the Prize for Today Change Progress at the United Nations climate-change conference in Paris in 2015, due to innovation and remarkable achievements in air-pollution control. The air quality of urban areas has gradually improved, but a shift in the characteristics of the main air pollutants has been unclear. The scientific research on the Lanzhou Blue phenomenon has not been published, which has been of no help to the other polluted Chinese cities in terms of learning about the Lanzhou experience of air-pollution prevention.

In this study, interannual, annual, and diurnal variations of six criterion air pollutants (PM_2.5, PM₁₀, SO₂, NO₂, CO, and O₃) in Lanzhou from 2013 to 2016 were analyzed using air-quality data from the China National Environmental Monitoring Centre. The interannual variation trends of the air pollutants were fitted using unary linear regression (ULR) to quantify variation of the pollutants. The study provides deep analysis of the variation of air-pollution characteristics in urban areas of Lanzhou, along with the shift of polluted air quality from the most to the less, which provides a basis for the formulation of future urban air-pollution control measures in central or eastern China, especially on the North China Plain.

2 Data and methods 2.1 Air-quality data

The air-quality data of urban areas and the outer suburbs of Lanzhou City used in this study were from state-controlled sites of the China National Environmental Monitoring Center (http://113.108.142.147:20035/emcpublish/). The applicable rules of state-controlled air-monitoring sites, the instruments measuring gaseous and particulate matter pollutants, the quality assurance measures, and the control of data were described in detail by Zhao et al. (2016) .

A automated monitoring systems were installed to measure concentrations of gaseous pollutants (SO₂, NO₂, CO, and O₃) at each site. Each monitoring system of gaseous pollutants is made up of a sample unit, a calibration unit, an analytical instrument, and a data-collection and transport unit. SO₂, NO₂, and O₃ are measured by the ultraviolet fluorescence method (TEI, Model 43i from Thermo Fisher Scientific Inc., USA), the chemiluminescence method (TEI Model 42i from Thermo Fisher Scientific Inc., USA), and the UV-spectrophotometry method (TEI model 49i from Thermo Fisher Scientific Inc., USA), respectively. CO is measured using the nondispersive infrared-absorption method and the gas-filter correlation/infrared-absorption method (TEI, Model 48i from Thermo Fisher Scientific Inc., USA). The PM_2.5 and PM₁₀ continuous monitoring system consists of the sample-collection unit, the sample-measurement unit, the data-collection and transport unit, and accessory equipment. The micro-oscillating balance method (TEOM from Rupprecht & Patashnick Co., Inc., USA) and the β-absorption method (BAM 1,020 from Met One Instrument Inc., USA) are used to measure PM _2.5 and PM₁₀.

The interannual variation trends of six criterion air pollutants at each state-controlled site of Lanzhou from 2013 to 2016 were fitted using unary linear regression (ULR) for domestic- (November to March of the next year) and no-domestic-heating seasons (April to October). The equation of unary linear regression can be expressed as follows:

$y = a + bx$

(1)

where y is a specific air pollutant, such as PM_2.5, PM₁₀, SO₂, NO₂, CO, or O₃; x is the number of years; and b is the rate of change for the specific air pollutant concentrations from 2013 to 2016.

2.2 Monitoring sites

Lanzhou is a typical valley city in an arid area, with a population of approximately 3.5 million, annual rainfall of 327.7 mm, and mean temperature of 9.3 °C. It is located in a long valley running mainly east to west with a length of about 30 km, maximum width of 8 km, and depth of 200–600 m. Due to the effect of the valley terrain, mean surface wind speed is only 0.3 m/s in winter; and the monthly calm frequency reaches as high as 81% (Wang et al., 1999 ). Furthermore, a lot of industries are based in the city, including an oil refinery, petrochemical industries, machinery and metallurgical industries, textile mills, food-processing centers, cement manufacturing, coal mining, rubber processing, electrical power generation, medical industries, lead and zinc mining and smelting, and fertilizer plants (Zhang XW et al., 2012 ). The air pollutants emitted from the industries were difficult to disperse due to weak winds and stable stratification inhibiting the turbulent diffusion aspect in urban Lanzhou, with the aspect ratio of the valley (depth versus width) of 0.07 (Chu et al., 2008 ).

The five state-controlled monitoring sites of Lanzhou are shown in Figure 2: (1) LLBG (103.63°E, 36.10°N) in the Xigu District, an industrial area with an oil-refinery company, petrochemical plant, textile mill, and other small industries; (2) ZGYY (103.71°E, 36.07°N) in the Qilihe District, residential and commercial areas near the Lanzhou Petro-Chemical Machinery Plant, which produces large machines for oil production; (3) SWZPS (103.84°E, 36.07°N) in the Chengguan District, a residential area near the Lanzhou Institute of Biological Products, which performs research and manufacturing of biological products; (4) TLSJY (103.83°E, 36.04°N) in the Chengguan District, residential and commercial areas near a machinery industry; and (5) YZLDXQ (104.14°E, 35.94°N) in Yuzhong County of Lanzhou City, which can be used as rural background regions. The interannual variations of PM_2.5 concentrations for daytime (the white bar) and nighttime (the black bar) in winter (the narrow bar) and summer (the wide bar) at each monitoring site are also given in Figure 2. The column of PM_2.5 concentration in 2016 is different from other years in Figure 2, mainly due to the lack of PM_2.5 data in winter 2016.

3 Results and discussion 3.1 Variations in the air pollutants

Concentrations of six criterion air pollutants and the differences between domestic- and no-domestic-heating seasons at the five stated-controlled monitoring sites of Lanzhou City were calculated in Table 1 to know the contribution of domestic heating to urban air pollution. The concentrations of particulate matter, CO, SO₂, and NO₂ at all monitoring sites for the domestic-heating season were significantly higher than those for the no-domestic-heating season, due to the effect of coal burning for heating and poor diffusion conditions, while higher concentrations of ozone were found in the no-domestic-heating season. Furthermore, the monitoring site with the highest concentration of a specific pollutant among the five sites was different for domestic- and no-domestic-heating seasons, due to variation of pollution sources during the two seasons (Table 1). For example, for the no-domestic-heating season, concentrations of PM_2.5 and SO₂ at LLBG, PM₁₀ at ZGYY, CO and NO₂ at TLSJY, and CO at SWZPS were higher than the corresponding air pollutant at the other monitoring sites. These figures indicate that the air quality at LLBG and ZGYY was mainly affected by industry and construction fugitive dust, respectively, while the air pollution at TLSJY and SWZPS was mainly induced by motor vehicles. The mean SO₂ concentration at YZLDXQ in the domestic-heating season was 19.6 μg/m³, which was 1.4 times higher than that in the no-domestic-heating season, suggesting that the urban air quality in winter in northern China was largely affected by coal burning for heating.

The AQI, PM_2.5, PM₁₀, and especially SO₂ gradually decreased from 2013 to 2016 for urban and rural areas of Lanzhou as production of some industries has been strictly controlled since 2013 in urban Lanzhou (Lanzhou Statistical Yearbook, 2016). The daily mean PM₁₀ concentrations also decreased gradually during the study period, with rates of −0.0297 μg/(m ³·d) (urban areas) and −0.0303 μg/(m ³·d) (rural areas). However, CO and NO₂ concentrations had slightly increasing trends (Figure 3), especially in urban areas, due to the large number of motor vehicles—804.6 thousand in urban Lanzhou in 2015, with an annual growth rate of 30.87% (Lanzhou Statistical Yearbook, 2016). The differences inFigure 3 were obtained by subtracting AQI or the concentration of the specific pollutant in rural areas from that in urban areas. The NO₂ and SO₂ concentrations in urban areas were higher than those in rural areas, especially NO₂, suggesting the large effect of emissions from industries and motor vehicles on urban air quality. Additionally, it is more interesting to us that the O₃ concentration in rural areas was much higher than that in urban areas in 2016, with the maximum value larger than 200 μg/m³ in June 2016, which has seriously impacted crops and human health. With the lower NO₂ concentration in rural areas in June 2016, the high concentration of O₃ in rural areas may be mainly affected by VOCs (volatile organic compounds) emitted from biology in rural areas, as NO_x and VOCs were the main precursors of O₃ (NRC, 1991). Ozone pollution is likely to worsen in the future despite recent reductions in the emission of particulate matter (PM_2.5 and PM₁₀) and SO₂ (Wang et al., 2013 ). For example, recent severe air pollution in summer in Chengdu, located in the Sichuan Basin, was partly induced by high ozone concentrations (http://www.sc.gov.cn/ 10462/12771/2015/8/20/10349299.shtml). Therefore, fine particulate-matter (PM_2.5) and ozone pollution should be prevented collaboratively over China in the future to improve urban air quality effectively.

To quantify the increase or decrease of a specific air pollutant each year from 2013 to 2015, interannual variable rates of the annual mean AQI (Air Quality Index) and six criterion air pollutants at the five sites for domestic- and no-domestic-heating seasons were fitted by unary linear regression (Figure 4 and Table 2; figures for LLBG, TLSJY, SWZPS, and YZLDXQ are similar to Figure 4 and so were omitted here). The concentrations of PM_2.5, PM₁₀, and SO₂ were obviously reduced by a more significant decrease of PM_2.5 in daytime of summer (Figure 2), while the other three pollutants gradually increased (Table 2), which was in accordance with the above conclusions. The variations of the air pollutants in the no-domestic-heating season were more significant than those in the domestic-heating season, especially PM₁₀ in ZGYY decreasing 19.85 μg/m³ each year as building construction was mainly deployed in warm seasons (Figure 4 and Table 2). As an indicator of motor-vehicle pollution, CO concentration increased by 0.21 mg/m³ per year for the no-domestic-heating season at SWZPS, with a dense population, 3.2 times higher than that for the domestic-heating season, indicating vehicle-exhaust pollution was increasingly serious with the rapid increase in the number of motor vehicles (Lanzhou Statistical Yearbook, 2016). Additionally, the increase in ozone concentration for the domestic-heating season at YZLDXQ background station was the most significant among the five monitoring sites, while for the no-domestic-heating season, the increase was more obvious at LLBG and SWZPS due to low particulate-matter concentrations' leading to strong solar radiation during the no-domestic-heating season (Zhao et al., 2016 ).

Interannual variations of probability-distribution functions (PDF) of AQI (Air Quality Index) and the six air pollutants at the five monitoring sites from 2013 to 2015 were selected to better understand air-pollution characteristics in recent years. Due to the small differences among the sites of urban areas, the PDF variations of AQI and the pollutants at ZGYY were given in Figure 5 as representative of urban areas. The logarithmic scale was used in the vertical axes of AQI, PM_2.5, and PM₁₀, and the horizontal axes of CO, SO₂, NO₂, and O₃ to better display interannual differences of PDFs of AQI and the pollutants in Lanzhou. As can be seen from Figure 5, the peaks of probability-distribution functions of AQI, PM_2.5, and PM₁₀ changed less from 2013 to 2015; but the air-pollution episodes with high PM_2.5 or PM₁₀ significantly reduced in Lanzhou, findings inconsistent with the frequent severe haze pollution on the North China Plain (NCP) in recent winters (Batterman et al., 2016 ; Yan et al., 2016 ; Yang et al., 2016 ). For example, the results from Yang et al. (2016) indicated that observed winter haze days averaged over eastern China (105°E–122.5°E, 20°N–45°N) increased from 21 days in 1980 to 42 days in 2014 and from 22 to 30 days from 1985 to 2005. The pollution days with an SO₂ concentration higher than 10 μg/m³ obviously decreased from 2013 to 2015 in Lanzhou, as a series of strict emission-control measures were taken in recent years. However, high CO and NO₂ pollution events considerably increased in Lanzhou from 2013 to 2015, which may be partly due to less restriction of motor-vehicle exhaust emissions. Generally, the shapes of the probability-distribution functions (PDF) were comparable between urban and rural areas. However, the peaks of the PDFs were lower in rural areas than those in urban areas, due to more primary emission in urban areas. In addition, ozone pollution was increasingly severe in urban and rural areas of Lanzhou in recent years, which presented a large threat to human and ecological health (Chen et al., 2014 ). Ground-level ozone pollution and its precursors (NO_x and VOCs) should be strictly controlled over China in the future to protect crops and human health (Wang et al., 2017 ).

To better understand the difference in major air pollutants among the five monitoring sites of Lanzhou (Figure 2), mean annual variations of hourly mean AQI and six criterion air pollutants for the sites from 2013 to 2016 are shown in Figure 6. As can be seen from Figure 6, annual variations of AQI and PM₁₀ for all sites were bimodal, with two peaks in spring and winter months due to attacks of dust plumes from upstream Taklimakan and Gobi Deserts in spring and more anthropogenic pollutants such as domestic heating, as well as poor diffusion conditions, in winter. The chemical composition of PM₁₀ had a large difference between urban and rural areas due to varying sources. The ions of Ca²⁺ and Mg²⁺ of PM₁₀ were much higher in spring than in winter due to the impact of frequent dust plumes in spring, while the ions SO₄²⁻ and NO₃⁻, mainly originated from anthropogenic sources had much higher concentrations in winter than in summer. The concentrations of PM_2.5, CO, SO₂, and NO₂ had maximum values in the winter months as a result of more emissions, in combination with weak winds and a long-standing temperature inversion (Chen et al., 1994 ). However, winter ozone concentrations were the lowest of the year, contrasting with the other air pollutants; these results were consistent with those from previous studies (Zhao et al., 2016 ; Yin et al., 2017 ). In addition, concentrations of all pollutants were much lower at the rural background station YZLDXQ than at other sites, with the exception of ozone, while the AQI and particulate-matter concentrations (PM_2.5 and PM₁₀) at LLBG were almost always the highest among the five monitoring sites over the year, due to a number of industries in Xigu District of Lanzhou (Lanzhou Statistical Yearbook, 2016). The highest monthly SO₂ concentrations variously appeared at LLBG (June to November), TLSJY (January to March, December), and SWZPS (April to May). CO and NO₂ concentrations for no-domestic-heating seasons (April to October) at TLSJY and SWZPS were much higher than those at the other monitoring sites, as a result of higher population and more motor vehicles in the Chengguan District of urban Lanzhou. The ozone concentrations were comparable among the five sites from January to September.

The interannual differences of mean diurnal variations of AQI and six criterion air pollutants in urban and rural areas of Lanzhou from 2013 to 2015 are shown in Figures 7 (urban areas) and 8 (rural areas), respectively. The differences of corresponding concentrations between 2013 and 2015 are also given in the two figures. For urban areas, the diurnal variation trends of AQI and the pollutants were similar among the three years; but the reductions of AQI, SO₂, PM_2.5, PM₁₀, and PM_2.5—accounting for PM₁₀ in daytime (08:00–20:00)—were more significant than those at nighttime (21:00–07:00), especially PM_2.5 (Figure 7), indicating fine particulate-matter pollution induced by human activities was controlled effectively due to implementing some strict emission-restriction measures in recent years. Additionally, increasing trends of NO₂ and CO concentrations varied with time of day, with the maximum at morning and evening rush hours, especially for CO, suggesting that motor-vehicle pollution is getting worse as the number of vehicles increases rapidly. Ground-level ozone concentrations obviously increased after noon, with the maximum increase of 15 μg/m³ at 17:00 from 2013 to 2015, as solar radiation reaching the earth surface obviously increased, caused by decreases of ambient particulate matter, which was beneficial to formation of secondary pollutants (Wang et al., 2017 ). Unlike the findings in urban areas, in rural areas of Lanzhou, interannual difference of mean diurnal variation trends of the pollutants was large (Figure 8). For example, not only the concentrations of SO₂, PM_2.5, and PM₁₀ obviously decreased from 2013 to 2015, but the peak of diurnal variations of the pollutants shifted from 13:00 to 11:00 in rural areas as well. Furthermore, the difference of ozone concentration between 2013 and 2015 had a maximum at 19:00 in rural areas, about two hours ahead of that in urban areas, which may be related to regional transportation of secondary pollutants formed, as YZLDXQ was located to the southeast of the urban areas of Lanzhou (Figure 2).

3.2 Variation of air-quality level and primary air pollutants

Frequencies of air-quality level (AQL) and the primary air pollutant at the five monitoring sites of Lanzhou City and their annual and interannual variation from 2013 to 2016 are showed (only summer is given), respectively, in Figures 9, 10, and 11 to explain the Lanzhou Blue phenomenon better. The number of days with good air quality (Good) at LLBG, mainly concentrated in summer (Figure 10), was obviously less than those at other monitoring sites; extremely and seriously polluted days and the days with PM_2.5 as the primary air pollutant were much more than at other sites, as LLBG is located in the Xigu District, an industrial area with many heavy industries. The impact is especially noticed in winter, as a result of more local emissions combining with poor diffusion conditions (Figures 9 and 10). The number of days with PM_2.5 as the primary air pollutant at YZLDXQ was comparable to that of other monitoring sites; but it had 5 days more with O₃ as the primary air pollutant than the other sites, mainly concentrated in April, May, and June (Figure 10). The days with PM_2.5 as the primary pollutant mainly appeared in winter (December, January, and February), as more fine particles were emitted from coal combustion or biomass burning for heating (Zhang HF et al., 2012 ; Zhang JK et al., 2016 ). However, the days with PM₁₀ as the primary pollutant chiefly occurred in the spring months (March to May) in Lanzhou City due to frequent dust plumes' being transported from Taklimakan and Gobi Deserts in its upstream regions (Zhao et al., 2015 ).

The number of summer days with good air quality increased from 2013 to 2016 in urban and rural areas of Lanzhou, while the slightly and moderately polluted days considerably reduced in recent years. Additionally, the days with PM_2.5 as the primary air pollutant obviously decreased in summer, especially at the background station YZLDXQ. However, the days with ozone as the primary pollutant began to appear in Lanzhou in 2015 and 2016, especially in 2016, which may be related to the increase of the concentration of NO₂, a main chemical precursor for formation of ozone (Wang et al., 2017 ). The above analyses indicate that particulate-matter pollution became slight in formerly highly polluted Lanzhou City in recent years, while secondary pollution such as ozone was increasingly severe, especially in rural areas. Some synergetic measures in urban and rural areas of Lanzhou should be taken by the local government to control fine particulate-matter and ozone pollution simultaneously.

4 Conclusions

Air quality improved in Lanzhou in recent years due to implementation of a series of strict emission-control measures, as evidenced in frequent appearances of the phenomenon of Lanzhou Blue. To better understand whether the concentration of each air pollutant was decreased significantly under the background of Lanzhou Blue and to provide some suggestions for urban air-quality improvement in the near future, the interannual, annual, and diurnal variations of AQI (Air Quality Index) and six criterion air pollutants (PM_2.5, PM₁₀, CO, SO₂, NO₂, and O₃) and variations of air-quality levels and the primary air pollutant at the five state-controlled monitoring sites of Lanzhou from 2013 to 2016 were studied using air-quality data from the China National Environmental Monitoring Centre. Some main conclusions were obtained, as follow.

The AQI, PM_2.5, PM₁₀, and especially SO₂ gradually decreased from 2013 to 2016 for urban and rural areas of Lanzhou, while CO and NO₂ concentrations had slightly increasing trends, especially in urban areas. The pollution days with SO₂ concentration higher than 10 μg/m³ obviously decreased from 2013 to 2015 in Lanzhou. The extremely and seriously polluted days and the days with PM_2.5 as the primary air pollutant at LLBG were much more than at other sites. The variation of the air pollutants in the no-domestic-heating season were more significant than those in the domestic-heating season. The increase of ozone concentration for the domestic-heating season at the YZLDXQ background station was the most significant among the five monitoring sites; while for the no-domestic-heating season, the increase was more obvious at LLBG and SWZPS. The number of the days with O₃ as the primary air pollutant at YZLDXQ was 5 days more than at the other sites and mainly concentrated in April, May, and June. Ozone pollution was increasingly severe in urban and rural areas of Lanzhou in recent years.

The production of some industries has been strictly controlled since 2013 in urban Lanzhou, but the vehicle-exhaust pollution was increasingly serious with the rapid increase in the number of motor vehicles. In recent years, particulate-matter pollution became slight in formerly highly polluted Lanzhou City, while secondary pollution such as ozone became increasingly severe, especially in rural areas. Some synergetic measures in urban and rural areas of Lanzhou should be taken by the local government in the near future to control fine particulate matter (PM_2.5) and ozone pollution simultaneously. Throughout China, ground-level ozone pollution and its precursors (NO_x and VOCs) should be strictly controlled in the future to protect crops and human health.

Acknowledgments:

The study is supported by Natural Science Foundation of China (41605103), Youth Innovation Promotion Association, CAS (2017462), CAS "Light of West China" Program, Opening Project of Shanghai Key Laboratory of Atmospheric Particle Pollution and Prevention (LAP³) (FDLAP16005), and the Excellent Post-Doctoral Program (2016LH0020). The authors also would like to thank the Qingyue Opening Environmental Data Center (QOEDC) for providing air-quality data used in this study.