Future haze events in Beijing, China: When climate warms by 1.5 and 2.0°C

Liu, Cuiping; Zhang, Feng; Miao, Lijuan; Lei, Yadong; Yang, Quan Article — Published Version Future haze events in Beijing, China: When climate warms by 1.5 and 2.0°C International Journal of Climatology Provided in Cooperation with: Leibniz Institute of Agricultural Development in Transition Economies (IAMO), Halle (Saale) Suggested Citation: Liu, Cuiping; Zhang, Feng; Miao, Lijuan; Lei, Yadong; Yang, Quan (2020) : Future haze events in Beijing, China: When climate warms by 1.5 and 2.0°C, International Journal of Climatology, ISSN 1097-0088, Wiley, Chichester [u.a.], Vol. 40, Iss. 8, pp. 3689-3700, https://doi.org/10.1002/joc.6421 , https://rmets.onlinelibrary.wiley.com/doi/full/10.1002/joc.6421 This Version is available at: https://hdl.handle.net/10419/222439 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by/4.0/ RESEARCH ARTICLE Future haze events in Beijing, China: When climate warms by 1.5 and 2.0C Cuiping Liu 1 | Feng Zhang 1 | Lijuan Miao 2,3 | Yadong Lei 1 | Quan Yang 1 1 Key Laboratory of Meteorological Disaster, Ministry of Education, Nanjing University of Information Science and Technology, Nanjing, People's Republic of China 2 School of Geographical Sciences, Nanjing University of Information Science and Technology, Nanjing, People's Republic of China 3 Department of Structural Development of Farms and Rural Areas, Leibniz Institute of Agricultural Development in Transition Economies, Halle, Germany Correspondence Lijuan Miao, Nanjing University of Information Science and Technology Nanjing, People's Republic of China Email: [email protected] Funding information National Key R & D Program of China, Grant/Award Number: 2017YFA0603503 Abstract China is the world's second-largest economy, and its capital Beijing has been suffering from severe haze pollution in recent years. However, how the winter haze events in Beijing vary under different global warming scenarios is still open for debate. In order to analyse long-term winter haze characteristics in Beijing in the future, we have simulated haze events using the haze weather index (HWI) for the warming periods of 1.5 and 2.0C, based on 20 Coupled Model Intercomparison Project Phase 5 (CMIP5) models under two representative concentration pathways (RCP4.5 and RCP8.5). Our results indicate that 16 CMIP5 models have preferable performance in simulating the spatial pattern and occurrence frequency of winter haze events in Beijing. We highlight that in the 1.5 and 2.0C global warming period (2020s–2050s), Beijing will face a significant increasing trend (6–9% growth rate) in the occurrence of winter haze events compared with the reference period (1986–2005). The frequency of winter haze events under the RCP4.5 increases less than under the RCP8.5 in the 1.5C warming period but is closer to RCP8.5 in the 2.0C warming period. The increase of winter haze events with respect to natural factors in Beijing could be attributed to stronger atmospheric inversions, weaker East Asian winter monsoons, and a shallowing East Asian trough induced by global warming. Our results will provide scientific instructions for environmental departments to better face meteorological hazards, such as air pollution episodes, thereby improving the early warning mechanism system for global warming. KEYWORDS 1.5C, 2.0C, China, CMIP5, global warming, haze 1|INTRODUCTION Air pollution is a major environmental problem all over the world. Satellite-derived estimates suggest that 30% of the global population lives in regions above the World Health Organization (WHO) interim target one standard (35 μgm −3 ) for PM 2.5 in 2010–2012 (Brauer et al., 2012; van Donkelaar et al., 2015). According to the Global Burden of Diseases study, air pollution is the fifth-ranked global risk factor for human health, responsible for over 5.5 million premature deaths (Cohen et al., 2017). Fine particulate matter is one major air pollutant that endangers human health, degrades visibility, and indirectly affects the global climate by participating in the clouding Received: 29 April 2019 Revised: 28 October 2019 Accepted: 26 November 2019 Published on: 18 December 2019 DOI: 10.1002/joc.6421 This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. © 2019 The Authors. International Journal of Climatology published by John Wiley & Sons Ltd on behalf of the Royal Meteorological Society. Int J Climatol. 2020;40:3689–3700. wileyonlinelibrary.com/journal/joc 3689 and raining process (Tamara et al., 2014; Wang et al., 2014a; Li et al., 2016b; Broomandi et al., 2017). From a global perspective, there are several areas with high PM 2.5 concentrations (e.g., for example, East Asia, India, Bangladesh, the Middle East, and North Africa) (van Donkelaar et al., 2015; Marlier et al., 2016). As one of the main megacities in Asia, Beijing and its adjacent economic areas have suffered serious airborne pollution, which have been further catalysed by frequent haze events since 2012 (Chen and Wang, 2015; Chen et al., 2017). In January 2013, there was an extremely severe and persistent haze pollution over eastern China, affecting 30 cities and 800 million people over an area of 1.3 million km 2 and cancelling hundreds of flights (Huang et al., 2014; Wang et al., 2014b). Haze has become one of the most sever meteorological disaster, globally and regionally. In general, haze is an atmospheric obscuration caused by fine particulate from various sources under specific meteorological conditions (Baklanov et al., 2016). When the pollutant particles are directly discharged in the ground layer or converted in the atmosphere, and if the horizontal and vertical dispersion of pollutants is restrained, haze events occur. Low surface wind, high surface relative humidity, and a stable boundary layer can all restrain the dispersion of pollutants (Ding and Liu, 2013; Yin and Wang, 2017). Meteorological conditions not only determine the dispersion of the pollutants but also greatly control the formation of secondary aerosol precursors, including SO 2 , NO x ,andVOC s , which is an important part of haze formation in China (Huang et al., 2014; Pei and Yan, 2018). The important role of meteorological factors on air pollution can be proven by the extremely severe haze pollution over eastern China in January 2013, analysis of this so-called “airpocalypse”period found that it could be attributed to the combined effect of various kinds of meteorological factors (Zhang et al., 2013b; Wang et al., 2014a). The meteorological condition is just one of the important factors affecting haze events, the discharge of pollutants is the leading factor. However, when it comes to the positive or negative effects on haze events in the context of climate change, the response of relevant meteorological conditions should be focused on. These meteorological factors are changing under the impact of atmospheric circulation. Thus, it is crucial to understand the potential effects of relevant natural factors on haze, both for the purpose of environmental management and investigating the social consequences of global warming (Leung and Gustafson Jr, 2005; Jacob and Winner, 2009). Recently, the role of underlying climatic factors in association with haze events in regional weather conditions has been explored, and it is projected that these factors will influence the haze situation in the future (Niu et al., 2010; Zhang et al., 2015; Zou et al., 2017). By analysing the Coupled Model Intercomparison Project Phase 5 (CMIP5) datasets, there are positive and negative contributors among different climate factors for the haze pollution in China (Han et al., 2017). The decadal variability and climate change, including the weakened East Asian winter monsoon system and the associated decreased nearsurface wind speeds (Niu et al., 2010), the increase of relative humidity (Chen and Wang, 2015; Chen et al., 2017), and the melting of Arctic ice (Wang et al., 2015; Cai et al., 2017), may provide favourable external conditions for future haze. On the contrary, the projected increase in precipitation is expected to relieve the haze pollution (Tian et al., 2015; Li et al., 2018b). Changes of atmospheric circulation driven by global warming are expected to alter the natural conditions that control haze formation and elimination (Li et al., 2018b), but the magnitude and direction of this change are still unclear. However, how the haze situation in China will vary with the integrated change in climate is still open for debate. A series of research projects have explored future haze events with an underlying idea to construct a comprehensive meteorological index for haze events (Horton et al., 2014; Han et al., 2017; Yin et al., 2017). Air environment carrying capacity, which measures atmospheric capacity in transporting and dispersing pollutants into the atmosphere globally, provides a direct way to investigate the change in the haze pollution potential (Han et al., 2017). Accumulation and diffusion of pollutants are closely related to the local atmospheric stability (Tai et al., 2012). Furthermore, Horton et al. (2014) applied a modified version of the Atmospheric Stagnation Index (ASI) to realize the quantification of the global warming effect on atmospheric stagnation. In addition, Winter Haze Day using the surface visibility and relative humidity is proposed in Yin et al. (2017). Later, Cai et al. (2017) constructed a local haze weather index (HWI) using the temperature difference between upper and lower troposphere, lower meridional wind, and upper zonal wind. Only 28% of haze events were captured by a winter ASI, whereas 48% of the HWI >0 days are haze days in Beijing, and the correlation between PM 2.5 and the HWI can be up to 0.66. A heavy pollution event is often the result of the comprehensive effect of various kinds of meteorological factors, which was further proven by research on weather conditions for severe haze events in January 2013 (Zhang et al., 2013b). Therefore, consideration of different meteorological parameters will help us to sensitively explore the future haze situation. In this study, we aim to estimate the potential frequency of winter haze events under the global warming scenarios of 1.5 and 2.0C, mainly based on a reanalysis dataset and a CMIP5 dataset. Our results will provide scientific instruction for researchers in environmental 3690 LIU ET AL. departments to better face meteorological hazards, such as air pollution, thereby improving early warning mechanisms for global warming. 2|DATASETS AND METHODOLOGY 2.1 |Datasets 2.1.1 |NCEP reanalysis data Daily climate datasets from 1986 to 2005 including atmospheric temperature and wind vector were provided by the National Centers for Environmental Prediction (NCEP), available from the following website (https:// www.esrl.noaa.gov/psd/data/gridded/reanalysis/). They are characterized with a resolution of 2.5 ×2.5 (Kalnay et al., 1996). 2.1.2 |CMIP5 datasets Twenty CMIP5 models with different spatial resolutions were applied in our study (Table 1), including daily outputs of historical and future climate experiments, under two representative concentration pathways (RCP4.5 and RCP8.5). RCP4.5 and RCP8.5 are named after the concentrations that approximate the intended level of radiative forcing values of 4.5 and 8.5 Wm −2 , respectively, in the year 2100 relative to pre-industrial values. For the consistency of different datasets, we interpolated all of the 20 CMIP5 experiment outputs to the same resolution of 2.5 ×2.5as that of the NCEP reanalysis data. The detailed variables of the two datasets used in this study are provided in Table 2. 2.2 |Methodology 2.2.1 |Haze weather index In this study, we focus on the frequency of haze events during boreal winter (December, January, and February), as this period typically experiences the most severe haze events in Beijing (Niu et al., 2010; Ding and Liu, 2013). Winter haze occurrence in Beijing is accompanied by a complex combination of meteorological factors. Here, we involve three meteorological elements when calculating HWI: temperature differences of the upper and lower troposphere, meridional winds in 850 hPa, and zonal winds in 500 hPa over Beijing (Cai et al., 2017). The three TABLE 1 Basic information of 20 CMIP5 models ID Short name Institution 1 ACCESS1.0 Australian Community Climate and Earth-System Simulator, Australia 2 ACCESS1.3 Australian Community Climate and Earth-System Simulator, Australia 3 BCC-CSM1.1 Beijing Climate Center, China Meteorological Administration, China 4 CanESM2 Canadian Centre for Climate Modelling and Analysis, Canada 5 CMCC-CM Centro Euro-Mediterraneo per I Cambiamenti Climatici, Italy 6 CMCC-CMS Centro Euro-Mediterraneo per I Cambiamenti Climatici, Italy 7 CNRM-CM5 Centre National de Recherches Météorologiques Coupled Global Climate Model, France 8 CSIRO-Mk3.6.0 Commonwealth Scientific and Industrial Research Organization, Australia 9 GFDL-CM3 Geophysical Fluid Dynamics Laboratory, USA 10 GFDL-ESM2M Geophysical Fluid Dynamics Laboratory, USA 11 HadGEM2-CC Met Office Hadley Centre, UK 12 INM-CM4 Institute for Numerical Mathematics, Russia 13 IPSL-CM5B-LR Institute Pierre-Simon Laplace, France 14 MIROC5 Model for Interdisciplinary Research on Climate, Japan 15 MIROC-ESM Model for Interdisciplinary Research on Climate, Japan 16 MIROC-ESM-CHEM Model for Interdisciplinary Research on Climate, Japan 17 MPI-ESM-LR Max Planck Institute for Meteorology (MPI-M), Germany 18 MPI-ESM-MR Max Planck Institute for Meteorology (MPI-M), Germany 19 MRI-CGCM3 Meteorological Research Institute, Japan 20 NorESM1-M Norwegian Climate Centre, Norway LIU ET AL.3691 meteorological elements are normalized to ΔT, V850, and U500: The ΔTconsiders the anomaly of temperature difference between the upper and lower troposphere and is expressed in the following formula: ΔT=T850 −T250 ðÞ−avg T850his −T250his ðÞ std T850his −T250his ðÞ :ð1Þ In this formula, T 850 denotes the area-averaged air temperature (32.5–45N, 112.5–132.5E) at 850 hPa and T 250 is the area-averaged air temperature (37.5–45N, 122.5–137.5E) at 250 hPa; T 850_his and T 250_his are the same as T 850 and T 250 , but for the historical period (1986–2005); avg indicates a time average function and std is a time standard deviation function. When ΔT>0, it means that vertical temperature anomalies of warming in the lower layer and cooling in the upper layer, which can strengthen the atmospheric stability and restrain the vertical diffusion of pollutants. The V850 considers the anomaly of meridional wind in the lower troposphere and is expressed in the following formula: V850=v850 −avg v850his ðÞ std v850his ðÞ :ð2Þ In this formula, v 850 indicates the area-averaged meridional wind speed (30–47.5N, 115–130E) at 850 hPa. For other parameters in Equation (2), please refer to Equation (1). When V850 >0, in the lower troposphere (850 hPa, shown in Figure 1c), anomalous southerly winds in eastern China can lead to the pollutants to remain confined in the local environment under the background of northwest prevailing winds in winter. The U500 considers the anomaly of zonal wind difference between north and south Beijing in the middle troposphere and is expressed as follows: U500=u500north −u500south ðÞ−avg u500northhis −u500southhis ðÞ std u500northhis −u500southhis ðÞ ð3Þ In this formula, u 500_north is the area-averaged zonal winds at 500 hPa in north Beijing (42.5–52.5N, 110–137.5E)，and u 500_south is the area-averaged zonal winds at 500 hPa in south Beijing (27.5–37.5N, 110–137.5E). Similarly, for other instructions, please refer to Equation (1). When U500 >0, an anomalous wind field of westerly wind strengthening in the north and weakening in the south of Beijing, bringing about an unfavourable situation for convection development and horizontal dispersion. These three anomalies are all referenced to the historical daily climatology and normalized by the historical standard deviation to facilitate comparison between the historical and future climate. Finally, three standardized indexes are summed and then normalized to get the HWI. HWI= ΔT+V850+U500ðÞ−avg ΔT+V850+U500ðÞ std ΔT+V850+U500ðÞ ð4Þ Positive HWI is advantageous for haze events and indicates high potential for their occurrence, whereas negative HWI is disadvantageous for haze events and indicates an unlikeliness for their occurrence. It should be pointed out that a HWI >0 shows weather conditions conducive to pollutants accumulating in the local environment. As for HWI >1, it indicates a higher severe haze pollution potential, but it fails to capture as many days as possible with haze occurrence potential. For convenience, we describe a HWI >0 day as a “haze event” and a HWI >1 day as a “severe event”in our study. 2.2.2 |Definition of 1.5 and 2.0C scenarios In 2015, the 21st Conference of the Parties to the Paris Accords agreed to take steps toward limiting the global mean annual surface air temperature increase to well below 2.0C above pre-industrial levels and to pursue efforts toward a target of 1.5C (Schleussner et al., 2016). As part of an ambitious global plan, more integrated analysis on the effects of limiting global warming to 1.5 and 2.0Cisrequired(Lang and Sui, 2012; Li et al., 2018b; Shi et al., 2018). In order to be consistent with the 20-year reference period (1986–2005), a moving average method is used to calculate the 20-year-averaged global mean surface annual temperature under the RCP4.5 and RCP8.5 (Schleussner et al., 2016; Shi et al., 2018). We further determine the warming periods TABLE 2 All variables using in the research Datasets/ variables Time period Surface 850 hPa 500 hPa 250 hPa NCEP (1986–2005) Surface temperature Air temperature, wind vector Wind vector Air temperature CMIP5 (1986–2005) (2005–2100) Surface temperature, surface level pressure Air temperature, wind vector, geopotential height Wind vector, geopotential height Air temperature 3692 LIU ET AL. with 1.5 and 2.0C higher than the pre-industrial value (1850–1900). Accounting for the model discrepancy on the start time of the historical experiment, the warming levels are derived relative to the reference period (1986–2005) instead of the pre-industrial period (1850–1900) (Schleussner et al., 2016). The reference period is 0.6C warmer than the preindustrial level (IPCC, 2013), which translates the global warming targets of 1.5 and 2C as warmings of 0.9 and 1.4C relative to the reference period level, respectively. 3|RESULTS 3.1 |Haze events simulation and model selection We evaluate performances of the 20 CMIP5 models in simulating the haze events (HWI >0) using Taylor diagrams (Figure 1). This method provides a concise statistical summary in a single diagram on how well two patterns match each other based on their correlation and the normalized standard deviation (Taylor, 2001). The anomalies of HWI >0 mean the meteorological conditions favourable to the occurrence of hazy pollution. Vertical temperature anomalies of warming in the lower layer and cooling in the upper layer shown in Figure 1a, can strengthen the atmospheric stability. The middle troposphere (500 hPa, shown in Figure 1b) displays an anomalous wind field of westerly wind strengthening in the north and weakening in the south of Beijing. In the lower troposphere (850 hPa, shown in Figure 1c), anomalous southerly winds in eastern China can lead to the pollutants to remain confined in Beijing area under the background of northwest prevailing winds in winter. Figure 1d–f displays the correlation coefficient and normalized standard deviation between NCEP datasets Pressure (hPa) (a) (d) (b) (e) (c) (f) FIGURE 1 Performance of CMIP5 models in simulating meteorological anomalies (HWI >0) during 1986–2005 period. Based on NCEP datasets: (a) 40N vertical section distribution anomaly of atmospheric temperature with HWI >0; (b) 500 hPa wind anomaly with HWI >0, shading indicates zonal flow; (c) 850 hPa wind anomaly with HWI >0, shading indicates meridional flow; (d)–(f) the correlation coefficients (the arc) and standard deviation (the x-axis) between each of the CMIP5 models and the NCEP datasets of the three meteorological anomalies using Taylor diagrams LIU ET AL.3693 and the CMIP5 models in the meteorological anomalies when HWI >0, using Taylor diagrams. Apparently, simultaneous meteorological patterns are shown by the NCEP datasets and the 20 CMIP5 models. The normalized standard deviations of CMIP5 models are distributed between 0.75 and 1.25, except for the model ACCESS1.0 and IPSL-CM5B-LR. The correlation coefficients from the other 18 models are higher and can reach 0.95 at a 99% confidence level, while only ACCESS1.0 and IPSLCM5B-LR own relatively low correlations. Thus, we remove ACCESS1.0 and IPSL-CM5B-LR from the 20 CMIP5 models for the purpose of decreasing error bias and keeping better consistency among the models. The remaining 18 models are further evaluated in simulating the occurrence frequency of HWI >0in Figure 2, as the second reference for model selection. It is clear that the models differ substantially in the frequency of favourable weather conditions from 1986 to 2005, when HWI >0. The frequency exhibits a wide range, from 866 to 947 days for HWI >0 and 288 to 326 days for HWI >1. By comparing with the NCEP datasets, we excluded the models named GFDL-CM3 and MPI-ESM- LR with the worst performance in simulating frequency of HWI >0 and HWI >1. As stated above, 16 CMIP5 models survived after two rounds of selection and proved to have a better ability in simulating weather conditions favourable to haze events. 3.2 |Climate change and haze events in the 1.5 and 2.0C warming scenarios Figure 3 shows changes in global mean surface temperature under RCP4.5 and RCP8.5 from 1986 to 2080. A good agreement in global mean surface temperature is shown between NCEP datasets and the ensemble mean from the selected 16 models, for the reference period 1986–2005. It illustrates that the projected difference of the global mean surface annual temperature under RCP4.5 and RCP8.5 is little before the 2020s, but gradually increases after the 2020s in Figure 3. The 1.5C warming targets would occur during the periods of 2019–2039 under RCP4.5, and during the 2015–2035 under RCP8.5. The 2.0C threshold of global warming would be reached in 2041–2061 and 2028–2048 under the RCP4.5 and RCP8.5, respectively. Compared with RCP4.5, a higher carbon emission pathway (RCP8.5) could yield a faster warming rate that crossing the 1.5 and 2.0C thresholds earlier. As shown above, the projection of the multi-models ensemble mean has reliable capability in simulating haze events for future warming scenarios. For the next step, we calculate the HWI for the warming periods to make a detailed statistical analysis of haze events (HWI >0) and severe haze events (HWI >1) (Figure 4). The statistical results of the multi-models ensemble mean for HWI >0 and HWI >1 in the reference period are consistent with these in the NCEP datasets, with a deviation rate of less than 5%. The frequency of HWI >0 in the reference period is lower compared to the NCEP datasets (916 days vs. 931 days), while that of HWI >1 is higher than the NCEP datasets (301 days vs. 288 days). Frequency (day) FIGURE 2 Days of weather conditions favourable to haze events in Beijing during boreal winter from 1986 to 2005. The blue column indicates days of HWI >0; the orange column indicates days of HWI >1. The black dashed lines show the reference days of HWI >0 (931 days) and HWI >1 (288 days) derived from NCEP 2.0°C GMT (°C) 1.5°C FIGURE 3 Changes in global mean surface temperature under RCP4.5 and RCP8.5 from 1986 to 2081. Light blue (light red) solid line: the 16-models-mean outputs under RCP4.5 (RCP8.5). Blue (red) solid line: The 20-year slipping averaged 16-models- mean outputs under RCP4.5 (RCP8.5). Translucent shading indicates the dispersion of one deviation from multi-model results. Horizontal reference lines in orange indicate the global warming cases of 1.5 and 2.0C 3694 LIU ET AL. Compared with the reference period, Beijing will experience more frequent haze days under the warming period. Under RCP4.5, the frequency of HWI >0 increases 53 days at a rate of 5.7% in the 1.5C warming period and 80 days at a rate of 8.7% in the 2.0C warming period. Under RCP8.5, it increases 78 days at a rate of 8.5% in the 1.5C warming scenario and 82 days at a rate of 8.9% in the 2.0C warming scenario. In the additional 0.5C warming scenario, there are only 4 days increasing in frequency with HWI >0 (Figure 4a), which is negligible. As for the severe haze events when HWI >1 (Figure 4b) under RCP4.5, the rate of frequency increasing can be up to 13.9% (42 days) and 19.9% (60 days) in the 1.5 and 2.0C warming periods, correspondingly, twice as much as for the haze events HWI >0. Similarly, the increasing frequency of HWI >1 under RCP8.5 is 60 days (19.9%) during the 1.5C warming period and 62 days (20.6%) during the 2.0C warming period. Only 2 days with HWI >1 under RCP8.5 increased in the additional 0.5C warming period. It is worth noting that the frequency of haze events under RCP8.5 is higher than RCP4.5 in the 1.5C warming period, but less than RCP4.5 in the additional 0.5C period, both for HWI >0 and HWI >1. 3.3 |Effects of possible mechanisms on haze events changes Our study shows a significant enhanced trend in haze days and severe haze days. The increasing frequency of haze event in the terms of meteorological factors is consistent with the mean state changes affecting the Beijing region. Figure 5 displays the changes of sea surface pressure and wind vector fields in the lower troposphere between the historical period and the warming periods under RCP4.5 and RCP8.5. A strengthened positive phase of Arctic Oscillation is displayed, which means a low in the Arctic area and two highs in the regions of the Pacific and Atlantic, respectively. Under this circumstance, cold air is confined within the high latitudes; a warmer land and a colder ocean lead to a weakening land-sea thermal discrepancy, resulting in weakening East Asian winter monsoons (Niu et al., 2010; Yin and Wang, 2017). In addition, East Asia is covered by widespread anomalous southerly winds, which also verifies the above analysis. The weaker north-east wind from the high latitude fails to blow the pollutants over Beijing away, hampering the horizontal diffusion of particulate matter. More stable atmospheric stratification, caused by surface warming, may inhibit the vertical diffusion of pollutants. The difference between Figure 5a,b explains why more HWI >0 (>1) days are shown under RCP8.5 in the 1.5C warming periods. As for the 2.0C warming periods, the similar intensity of Arctic Oscillation displayed in Figure 5c,d is one possible factor for no obvious HWI >0(>1) differences between the two scenarios. This is supported by a circulation field in the middle layer of the troposphere. Figure 6 shows an anomaly of the cyclonic circulation over the East Asia. In addition, a strong and obvious high anomaly controls mainland China, especially East Asia, implying the weakness of the East Asian trough during the future warming periods. The shallowing East Asian trough brings less cold and dry air to the Beijing area, thus favours the formation and maintenance of haze events (Chen and Wang, 2015; Chen et al., 2017). 4|DISCUSSION In this study, we applied a meteorological based index for the purpose of evaluating the haze events in Beijing (a) (b) FIGURE 4 Frequency difference of HWI >0(>1) (days) in warming periods relative to the reference period: (a) counts the days of HWI >0 in the historical periods and in the 1.5 and 2.0C scenarios; (b) corresponds to (a), but for HWI >1. The box whisker plots show the minimum, 25th, 50th, 75th, and maximum intervals of the CMIP5 experimental results. The green dots represent the results of the 16-models-mean. The black dot donates the referenced results using the NCEP reanalysis dataset LIU ET AL.3695 under the 1.5 and 2.0C global warming scenarios. The HWI considers three meteorological factors affecting the accumulation and diffusion of pollutants and is specifically designed for capturing Beijing haze events (Cai et al., 2017). All relevant variables can be simulated and provided based on the outputs of CMIP5 models. By contrast, application of other indexes is limited considering the data availability. For example, air environment carrying capacity involves variables, like boundary layer height, that are not provided in CMIP5 simulations (Horton et al., 2014; Kang et al., 2016; Han et al., 2017). The index of winter haze days based on visibility and relative humidity from ground observations is usually adopted by investigating the relationship between winter haze and climate factors in other studies as well (Li et al., 2016a; Zhang et al., 2016; Pei and Yan, 2018). However, those indexes are not appropriate when it comes to future winter haze under the background of global warming. HWI is a reliable and computable index for predicting the haze events considering weather conditions. The present results of increased frequency of haze events in Beijing in winter under the global warming background are consistent with recent relevant studies (Cai et al., 2017; Han et al., 2017; Pei and Yan, 2018; Chen et al., 2019). However, for the first time, our research provides quantitative assessment of haze frequency in Beijing at different warming scenarios of 1.5 and 2.0C under RCP4.5 and RCP8.5. The average frequency of winter haze events varies from 45 days in the reference period to 48 days during the 1.5C warming periods and 50 days during 2.0C warming periods under RCP4.5. Under RCP8.5, the winter average frequency in the 1.5C warming level increases to 50 days, while in the 2.0C warming level, it will not rise anymore. Many studies on the effects of the 2.0C warming highlight that there are substantial differences between 1.5 and 2.0C 0.5 m.s–1 0.5 m.s–1 0.5 m.s–1 0.5 m.s–1 (a) (c) (d) (b) FIGURE 5 The simulated 16-models-mean anomalies of sea level pressure and 850 hPa winds vectors in boreal winter. Mean changes of 1.5C warming periods and 2.0C warming periods relative to the reference periods under RCP4.5 and RCP8.5 are shown in (a) and (b), (c) and (d), respectively. HIS represents the period of 1986–2005, RCP4.5(1.5) and RCP8.5(1.5) represent the periods of 2019–2038 and 2015–2034, RCP4.5(2.0) and RCP8.5(2.0) represent 2041–2060 and 2028–2047. White dots represent the area where the sea level pressure changes are significant at the 95% confidence level based on a student t-test 3696 LIU ET AL.