Search

Beijing has suffered a series of poor air quality ("haze") episodes which have damaged human health and economic growth. Beijing's haze is generally driven by pollutant‐trapping meteorological conditions, but climate model projections differ on how these conditions will evolve under forcing. Such differences are driven in part by (1) disagreements over which meteorological conditions matter most and (2) an undersampling of internal variability in projections. To resolve these differences, we show that Beijing haze is associated with anticyclonic circulation, thereby linking multiple meteorological ingredients to a single process well simulated by models. We use this to inform our assessment of future haze risks in both a multimodel ensemble and a single‐model large ensemble, partitioning model uncertainty and internal variability. We estimate that forcing increases haze‐favorable conditions, but internal variability significantly affects these projections, emphasizing the importance of fully sampling sources of uncertainty to ensure accurate projections of haze risks. Plain Language Summary: Episodes of poor air quality ("haze") in Beijing, China have been linked to cardiovascular and respiratory disease, damage to crops, and decreases in economic growth. These episodes tend to occur when moist, still air lets pollutants accumulate. Human‐caused climate change may affect the weather conditions that cause haze, but that effect is uncertain due to how scientists characterize haze‐favorable weather, how such weather links to actual haze occurrence, and how those features change in climate model experiments that neither simulate pollutants explicitly nor represent the full distribution of climate variability ("internal variability"). Here we show that different weather conditions that promote haze in Beijing are linked to the same large‐scale wind patterns, putting those conditions in the context of a single physical process simulated by climate models. When evaluated against this observed benchmark, we can confidently use multiple climate model ensembles to estimate how climate change affects haze‐favorable conditions while also addressing uncertainties due to model choices and internal variability. We show that climate change will make haze‐favorable conditions over Beijing more frequent, but that internal variability can generate large uncertainties in these projections, demonstrating the importance of fully sampling climate variability to constrain the risks of climate‐induced health impacts. Key Points: Poor air quality in Beijing is linked to stagnant meteorology at the surface and anticyclonic circulation that promotes such conditionsClimate models project increases in atmospheric conditions favorable to pollutant accumulation throughout the 21st centuryInternal variability strongly affects such projections, implying that previous projections have undersampled initial‐condition uncertainty [ABSTRACT FROM AUTHOR]

High‐impact poor air quality events, such as Beijing's so‐called "Airpocalypse" in January 2013, demonstrate that short‐lived poor air quality events can have significant effects on health and economic vitality. Poor air quality events result from the combination of the emission of pollutants and meteorological conditions favorable to their accumulation, which include limited scavenging, dispersion, and ventilation. The unprecedented nature of events such as the 2013 Airpocalypse, in conjunction with our nonstationary climate, motivate an assessment of whether climate change has altered the meteorological conditions conducive to poor winter air quality in Beijing. Using three indices designed to quantify the meteorological conditions that support poor air quality and drawing on the attribution methods of Diffenbaugh et al. (2017, https://doi.org/10.1073/pnas.1618082114), we assess (i) the contribution of observed trends to the magnitude of events, (ii) the contribution of observed trends to the probability of events, (iii) the return interval of events in the observational record, preindustrial model‐simulated climate and historical model‐simulated climate, (iv) the probability of the observed trend in the preindustrial and historical model‐simulated climates, and (v) the relative influences of anthropogenic forcing and natural variability on the observed trend. We find that anthropogenic influence has had a small effect on the probability of the January 2013 event in all three indices but has increased the probability of a long‐term positive trend in two out of three indices. This work provides a framework for both further understanding the role of climate change in air quality and expanding the scope of event attribution. Key Points: Extreme air quality conditions present in Beijing during January 2013 represented by three indices of air quality meteorologyNatural variability found to have a larger influence on extreme air quality events than anthropogenic forcingAnthropogenic forcing drove long‐term positive trends in two out of three indices, despite small effect on individual events [ABSTRACT FROM AUTHOR]