Supplementary Information Rapid formation and evolution of an extreme haze episode in Northern China during winter 1 Yele Sun 1,*, Chen Chen 1,, Yingjie Zhang 1,, Weiqi Xu 1,3, Libo Zhou 1, Xueling Cheng 1, Haitao Zheng 1,, Dongsheng Ji 1, Jie Li 1, Xiao Tang 1, Pingqing Fu 1, Zifa Wang 1 1 State Key Laboratory of Atmospheric Boundary Layer Physics and Atmospheric Chemistry, Institute of Atmospheric Physics, Chinese Academy of Sciences, Beijing 19, China Center for Excellence in Urban Atmospheric Environment, Institute of Urban Environment, Chinese Academy of Sciences, Xiamen 311, China 3 University of Chinese Academy of Sciences, Beijing 19, China College of Applied Meteorology, Nanjing University of Information Science and Technology, Nanjing 1, China School of Atmospheric Physics, Nanjing University of Information Science and Technology, Nanjing 1, China Key Laboratory of Environmental Optics & Technology, Anhui Institute of Optics and Fine Mechanics, Chinese Academy of Sciences, Hefei 331, China *Corresponding author email: sunyele@mail.iap.ac.cn; Phone: +-1--1 1
Severe Haze Episodes in 1. Two severe haze episodes were observed during the same winter period in 1, each lasting approximately two days. The evolution of meteorological parameters and aerosol composition was classified into seven episodes shown in Fig. S. Similar to the 1 haze episode, the formation of these two episodes (F1 and F, Fig. S) was initiated by a change in air masses to the south, an increase of RH and a decrease of T. But, the evolution of the first haze episode (Ep and Ep3) was driven by southerly and southwesterly winds through the entire vertical layer (Fig. Sa). The ground wind speed was low (< m s -1 ) and higher wind speeds were observed at a higher altitude (> m s -1 at m). Under these conditions, the accumulation of secondary aerosol species was much slower than that in 1. For example, it took nearly a day for sulfate and nitrate to reach approximately µg m -3, which was less than half the 1 levels. The chemical composition of PM 1 during Ep was dominated by secondary aerosols, accounting for % on average. Sulfate, nitrate, and SOA were roughly equivalent contributing 13% to 1%. As indicated in Figure 3, the evolution of this episode was influenced by mountain-valley breezes at : on November (M1 in Fig. S) that substantially reduced secondary aerosol species. After the mountain-valley breezes disappeared and the wind direction switched back to southwesterly, aerosol species started to increase (Ep3) and the PM 1 mass concentration increased from 3 to 19 µg m -3. Aerosol composition showed significant changes, with secondary aerosol species contributing % of PM 1 on average. In particular, sulfate increased and led to an enhancement in sulfate contribution from 1% to %. During this episode, variations in primary and secondary aerosol species were quite different. While the evolution of secondary aerosol species was principally driven by meteorological variations on a regional scale, primary aerosol species showed strong diurnal variations that appeared to be independent of meteorological conditions. Compared to the OA composition in the severe haze episode of 1, POA was more significant than SOA in 1, and accounted for 7% and % during Ep and Ep3, respectively. CCOA was the largest
POA component, on average contributing % and 3% of the total OA. This suggests that coal combustion emissions were the major source of OA during these two episodes. The evolution of the second haze episode was similar to the first one. While the average mass concentration of PM 1 increased from 9 µg m -3 in the early formation stage (Ep) to 19 µg m -3 during the fog stage (Ep), sulfate increased the most from 1% to %. Secondary aerosol species dominated aerosol composition during both episodes, accounting for 7% and 7%, respectively. This highlights the dominant role of regional transport in the formation of severe haze episodes. But, we also noticed the considerable impact of local primary sources on OA, with contributions ranging from 3 % during Ep, Ep3, Ep and Ep. Among POA components, CCOA was dominant, contributing 3% of OA. This further illustrates the important role of coal combustion emissions in the formation of winter haze. 3
Table S1. A summary of average aerosol species concentrations, optical properties, gaseous species, and meteorological parameters for the six episodes in Fig. 1. Ep1 Ep Ep3 Ep Ep Ep Species (µg m -3 ) SOA.3. 7. 7.1 1.19.99 SO 1.3.71.7 7.9 9.11 1. NO 3. 13.17..13 31.9.1 NH..7 1.33 13. 1.39.37 BC Sec.3..93 1. 1.9.3 POA 1.7 1.3 3. 77.9..7 Chl...1.3 11.1.3 BC pri..1.3 7. 7..11 BC.7 7.1 13.3 1.9 3..33 PM 1.99. 1. 19. 3..1 PM..1 131.3 7. 33.3.3 3.7 Optical properties Ext (M m -1 ) 37. 1.9 17.11 17. 93.1.71 Abs (M m -1 ).. 9.7 1. 171..3 SSA..9.9.93.9.91 Gas species (ppb) NO.79 9.7 7. 9.71 1. 3.3 NO 13.1..9 7.3 1.1 1. NO x 1.3. 13.7 3. 1.7 1.9 CO. 1.9.39 7..1. SO 3.19.7 3.31 1.7 13. 3.1 O 3 7. 3.7.7..3 1. Meteorological parameters RH, m (%) 3. 3.. 73.. 31. RH, 1m (%) 3.1.1 7.3 9. 97. 3.9 T, m ( C) -.7-3.3.1 3.1.11 3.7 T, 1m ( C) -.9 -. -1.1 3.99.9.93 WS, m (m s -1 ). 1.17 1..7.9.9 WS, 1m(m s-1).7. 1.9 1.77 1.7.
U * (m/s) 1... (a) m 1m 7m m 1m m m E u (m /s ). 1 E v (m /s ) E w (m /s ) b ext (M/m), PM... 1. 1... 3 1 (d) (e) 1: AM 11/ 1: AM 11/ 1: AM 11/7 1: AM 11/ 1: AM 11/9 1: AM 11/3 PM. (x ) 1: AM 1/1 b ext 1: AM 1/ Figure S1.Time series of (a) friction velocity and (b-d) turbulence kinetic energy (TKE) in horizontal (u and v) and vertical (w) wind components at seven heights. (e) shows the time series of PM. mass concentration and particle extinction coefficient (b ext ). u, v, w are the three components of velocity along x, y, z directions, respectively. The u, v, w are first averaged for min for the mean flows ( u, v, w), and then the turbulence is calculated as the departure of the instantaneous wind from the min mean, i.e., u u u =. The turbulence kinetic energy of motion (e) along x, y, z directions in the period domain less than min is calculated as e = u, e = v and e w w =, respectively, and the averages of e u, e v and e w for every min (ensemble u v mean for one case) are denoted E u, E v and E w, respectively, i.e., E u u =, E v v =, E w w =. A more detailed description of the calculation of TKE is given in Cheng et al. 1
Height (m) WD ( ) Height (m) 1 1 3 7 1 9 1 1 1 1 WS (m s -1 ) RH (%) (a) (d) (e) (f) Ep1 1 11% 1% 19% 33% M1 13% 1% 1% 3% M RH,m RH,1m T,m T,1m Ep 3 Ep3 19 LWC % 1% 1% % Ep 11 % 13% 1% 3% Ep 9 1% % 1% % Black: < 1. m s -1 Ep 19 % 1% 1% % Ep7 3% % % % W S E 1-3 1 m 1 m 7 m m 1 m m m 1 m 1 m 7 m m 1 m m m T ( o C) 1 1 Secondary: SOA Sulfate Nitrate Ammonium BC sec Primary: BC pri Chloride POA Wind Dir WS (m s -1 ) LWC (mg m -3 ) 1: AM 11/ M1 1: AM F1 11/ M 1: AM 11/ 1: AM 11/7 1: AM 11/ F 1: AM 11/9 1: AM 11/3 1: AM 1/1 Figure S. Evolution of (a, b) wind direction (WD), (c, d) wind speed (WS), (e) relative humidity (RH), temperature (T) and liquid water content (LWC), and (f) mass concentrations of aerosol chemical species in 1. Seven episodes (Ep1 Ep7) and two initial stages (F1 F) in the formation of the episodes are marked for further discussion. Pie charts show the average chemical composition of each episode. The horizontal dashed lines in (a) and indicate a height of m, and the dashed line in (d) indicates a wind speed of m s -1.
RH (%) WD ( ) NO, NO (ppb) 1 3 7 1 9 1 3 CO (ppm) Mass Conc. (µg m -3 ) 1 3 1 1 (a) (d) (e) 1: AM 11/ RH,m T,m SOA Sulfate Nitrate Ammonium Chloride POA BC Ep1 1: AM 11/ RH,1m T,1m Ep 1: AM 11/7 1: AM 11/ Ep3 1: AM 11/9 1: AM 11/3 1: AM 1/1 1: AM 1/ Figure S3. Time series of (a) relative humidity (RH) and temperature (T), wind speed (WS) and wind direction (WD), CO and SO, and (d) NO, NO, and O 3 from November to December, 1. (e) shows a stack plot of chemical composition of PM 1, single scattering albedo (SSA) and water vapor (H O) during the study period. Six episodes (Ep1 Ep) which are the same as in Figure are marked. Ep WS, m WS, 1m Ep CO SO NO NO O 3 HO (g m -3 ) Ep - - 1 3 1 3 1 WS (m s -1 ) T ( o C) SO (ppb) O3 (ppb).9..... 3. SSA SO (µg m -3 ) 3 f(x) = 1.9x + 13.9 r =.99 1 1 3 3 1 7 1 1 f(x) = 1.x +. r =.9 1 3 NO 3 (µg m -3 ) Figure S. Scatter plot of sulfate versus nitrate for episodes (identified by numbers in the figure) in 1 (red circles) and 1 (blue circles). A linear fit was performed on the three episodes with significant aqueous-phase production of sulfate, and separately on the remaining episodes. 7
(a) (d) (e) (f) Figure S. Footprint regions for air arriving at m during the six episodes (Ep1 - Ep) marked in Figures 1 and. The legend indicates the number concentrations of tracer particles. The maps were drawn by IGOR Pro (version.3.7., WaveMetrics, Inc., Oregon USA), http://www.wavemetrics.com/. The footprint region of each episode was determined using two-day backward simulations of the Lagrangian particle dispersion model FLEXPART driven by the meteorological field (spatial resolution = 1 km, time resolution = 1 hour). In this study, the meteorological simulations were carried out using the Weather Research and Forecast model version 3. (WRF) 3 with the National Centers for Environmental Prediction (NCEP) global reanalysis data as the initial and boundary conditions. There were two domains in the simulation with grid resolutions of 3 and 1 km, respectively, and vertical levels. In addition, the Yonsei University (YSU) boundary layer scheme, the Kain-Fritsch convective parametrization, the WSM3 microphysics scheme, the Dudhia shortwave scheme and the RRTM longwave scheme, were used in model simulations. Particle locations were then calculated with WRF-FLEXPART. In the simulations, 1, tracer particles were released from the site at two heights, m and m, respectively, and the model was run backwards to determine the source areas and transport pathways of air pollutants during the specified period. A larger number of tracers in a cell indicated a greater impact from surface emission sources. A more detailed evaluation of WRF-FLEXPART is given elsewhere 7.
(a) 1% (d) 1% Ep1 13% 3% % 13% 1% 3% 1% % 1% 3% D3 1% 7% 1% 1% 1% % % 1% Ep % 3% 3% 1% D 11% 9% 1% 1% COA 33% 1% 1% 7% 1% % % Ep3 CCOA 3% 1% SV-OOA HOA 1% % LV-OOA 37% CCOA HOA % COA 11% 3% LV-OOA SV-OOA1% % Ep NO 3 NH 1% BC % sec % BC pri % Chl 9% 1% 1% 9% 1% 1% D3 SO % % POA 7% % 37% 17% 1% SOA 1% 19% % Ep 3% % 13% % % % % % 7% 1% % 1% D % 3% % % Ep % % 1% 1% 11% 3% 13% % 3% Figure S. Average composition of organic aerosol during (a) six episodes (Ep1 Ep) and two events (D3, D) in 1. shows the average chemical composition of PM 1 for the D3 and D events in 1, and (d) shows the average composition during seven episodes (Ep1 Ep7) in 1. The episode information in 1 and 1 is marked in Figure and Figure 3, respectively. 9
RH (%) WD ( ) NO, NO (ppb) 1 3 7 1 9 CO (ppm) Mass Conc. (µg m -3 ) 1 1 1 1 1 F1 (a) (d) (e) Ep1 RH,m T,m Ep RH,1m T,1m Ep3 C1 Ep F Ep Ep C 1-1 3 1 WS (m s -1 ) WS,1m, WS,m CO SO 1 1 NO NO O 3 Ep7 SOA Sulfate Nitrate Ammonium Chloride POA BC T ( o C) SO (ppb) O3 (ppb) 1: AM 11/ 1: AM 11/ 1: AM 11/ 1: AM 11/7 1: AM 11/ 1: AM 11/9 1: AM 11/3 1: AM 1/1 Figure S7. Time series of (a) relative humidity (RH) and temperature (T), wind speed (WS) and wind direction (WD), CO and SO, and (d) NO, NO, and O 3 from November to December 1, 1. (e) shows a stack plot of chemical composition of PM 1 during the study period. Seven episodes (Ep1 Ep7) are identified and are included in the analysis for Figure 3. 1
Fraction of Total Signal. (a) HOA.3..1. CCOA...... COA..... (d) SV-OOA..... (d) LV-OOA.1.1.. 1 3 (a) HOA r HOA vs. NO x =. 3 r HOA vs. BC =. 1 1 1 CCOA r CCOA vs. CO =.7 3 1 r CCOA vs. Chl =.7 1 COA 3 1 1 (d) SV-OOA r SV-OOA vs. NO 3 =.3 1 (d) LV-OOA r LV-OOA vs. SNA =.1 1 1 7 9 1 11 1 13 1 111/3/1 11//1 11/7/1 11/9/1 1/1/1 m/z (amu) m/z (amu) Mass Conc. (µg m -3 ) BC (µg m -3 ) CO (ppm) 1 1 1 1 3 1 NO x (ppb) Chl (µg m -3 ) NO 3 (µg m -3 ) SNA (µg m -3 ) Figure S. Mass spectra (left panel) and time series (right panel) of five OA factors resolved from PMF analysis of organic aerosol spectra in 1. A comparison of OA factors with the external tracers is also shown in the right panel. Fraction of Total Signal (a) HOA.3..1. CCOA.... COA.....1 (d) SV-OOA.1.... (d) LV-OOA.1.1.. 1 3 m/z (amu) 7 9 1 (a) HOA r HOA vs. BC =. 1 1 CCOA r CCOA vs. CO =., r CCOA vs. NO x =.7 1 3 r CCOA vs. Chl =.7 3 1 1 COA 3 1 (d) SV-OOA r SV-OOA vs. NO 3 =.7 1 1 (d) LV-OOA r LV-OOA vs. SNA =. 3 1 11//1 11/7/1 11/9/1 1/1/1 m/z (amu) NO x (ppb) Mass Conc. (µg m -3 ) CO (ppm) 3 1 1 1 3 1 1 BC (µg m -3 ) Chl (µg m -3 ) NO 3 (µg m -3 ) SNA (µg m -3 ) Figure S9. Mass spectra (left panel) and time series (right panel) of five OA factors resolved from PMF analysis of organic aerosol spectra in 1. A comparison of OA factors with the external tracers is also shown in the right panel. 11
(a) Figure S1. Comparisons of the simulated and observed PM. concentrations from November, 1 to December 1, 1 at (a) Aoti centre in Beijing, Yongminglu in Tianjing, and Fenglongshan in Shijiazhuang. 1
References: 1 Cheng, X., Zeng, Q. C. & Hu, F. Characteristics of gusty wind disturbances and turbulent fluctuations in windy atmospheric boundary layer behind cold fronts. Journal of Geophysical Research: Atmospheres 11 (11). Stohl, A., Forster, C., Frank, A., Seibert, P. & Wotawa, G. Technical note: The Lagrangian particle dispersion model FLEXPART version.. Atmospheric Chemistry and Physics, 1-7 (). 3 Skamarock, W. C. et al. A description of the advanced research WRF version. (DTIC Document, ). Hong, S.-Y., Noh, Y. & Dudhia, J. A new vertical diffusion package with an explicit treatment of entrainment processes. Mon Weather Rev 13, 31-31 (). Kain, J. S. The Kain-Fritsch convective parameterization: an update. J Appl Meteorol 3, 17-11 (). Fast, J. D. & Easter, R. C. in 7th WRF User s Workshop, Boulder, CO, USA. (Citeseer). 7 de Foy, B. et al. Aerosol plume transport and transformation in high spectral resolution lidar measurements and WRF-Flexpart simulations during the MILAGRO Field Campaign. Atmos. Chem. Phys. 11, 33-33, doi:1.19/acp-11-33-11 (11). 13