Chinese SO 2 pollution over Europe – Part 2: Simulation of aerosol and cloud condensation nuclei formation

We report on sulfur dioxide (SO 2 ) induced formation of aerosols and cloud condensation nuclei in an SO 2 rich aged (9 days) pollution plume of Chinese origin, which we have detected at 5–7 km altitude during a research aircraft mission over the East Atlantic o ﬀ the West coast of Ireland. Building on our measurements of SO 2 and other 5 trace gases along with plume trajectory simulations, we have performed model simulations of SO 2 induced formation of gaseous sulfuric acid (GSA, H 2 SO 4 ) followed by GSA induced formation and growth of aerosol particles. We ﬁnd that e ﬃ cient photochemical SO 2 conversion to GSA took place in the plume followed by e ﬃ cient formation and growth of H 2 SO 4 -H 2 O aerosol particles. Most particles reached su ﬃ ciently large sizes 10 to act as cloud condensation nuclei whenever water vapor supersaturation exceeded 0.1–0.2%. As a consequence, smaller but more numerous cloud droplets are formed, which tend to increase the cloud albedo and to decrease the rainout e ﬃ ciency. The detected plume represents an interesting example of the environmental impact of long range transport of fossil fuel combustion generated SO 2 . 15


Introduction
Fossil fuel combustion represents the most important source of atmospheric sulfur dioxide (SO 2 ), a major air pollutant. Presently most atmospheric SO 2 is released in Europe and China from combustion of relatively sulfur rich coal (Lelieveld et al., 2001). Sulfur dioxide impacts the environment in several ways. It is toxic and after inhala-20 tion may cause severe adverse health effects (e.g. Sunyer et al., 2003;Venners et al., 2003;Longo et al., 2008). Moreover, it contributes to atmospheric acidity and acidinduced corrosion (e.g. Rodhe et al., 2002;Huang et al., 2008). In the atmosphere, SO 2 undergoes conversion to particulate sulfate and OH-induced gas-phase conversion to gaseous sulfuric acid (GSA, H 2 SO 4 ), (e.g. Menon and Saxena, 1998;Reiner Introduction Conclusions References Tables  Figures   Back  Close Full Screen / Esc

Printer-friendly Version
Interactive Discussion and upper troposphere, GSA may undergo binary (H 2 SO 4 -H 2 O) nucleation leading to new aerosol particles. These grow by condensation and coagulation and ultimately may become CCN (Seinfeld and Pandis, 2006). Consequently, atmospheric SO 2 promotes the formation of sulfuric acid-water aerosol particles which impact the radiation balance by absorbing and scattering sunlight and by prolonging the lifetime of clouds 5 (e.g. Ramanathan et al., 2001;Harshvardhan et al., 2002;Garrett et al., 2002;Andreae et al., 2005). After SO 2 release from ground level combustion sources substantial amounts of SO 2 may be transported to the middle and upper troposphere and thereby escape deposition at the surface and are less affected by removal via cloud processes. In the 10 middle and upper troposphere, SO 2 is removed preferably by OH induced gas-phase conversion to GSA which leads to a SO 2 lifetime of about 7-14 days depending on the efficiency of photochemical OH formation (e.g. Finlayson-Pitts and Pitts, 2000). Hence, the SO 2 lifetime is sufficiently long to allow SO 2 long-range transport. Therefore, SO 2 may impact aerosols and clouds thousands of kilometers away from its emis- 15 sion source. An interesting question is to what extent the important SO 2 source China impacts regional aerosols and clouds. Here we report on model investigations of SO 2 induced formation of GSA, aerosols and CCN in a major SO 2 rich pollution plume of Chinese origin, which we have detected over the East Atlantic (Fiedler et al., 2008). The model simulations are performed with an atmospheric chemistry and aerosol dy-mass spectrometry (CIMS) using a permanent calibration with isotopically labeled SO 2 . The objective was to probe an Asian pollution plume, which had been predicted before by the lagrangian particle dispersion model FLEXPART (Stohl et al., 2002(Stohl et al., , 2005. Between 10:00 and 11:00 UTC several pollution plumes were detected as particularly indicated by very markedly elevated SO 2 . A comprehensive description of the mea-10 surements especially in the light of other simultaneously measured trace gases can be found in Fiedler et al. (2008). Figure 1 shows the vertical distribution of the SO 2 mole fraction measured during that flight. Apart from a SO 2 pollution layer of American origin at approximately 1.5 km altitude, a deep layer of SO 2 has been detected in the altitude range from 5 to 7.5 km. 15 The atmospheric SO 2 background in this layer is about 150 pmol/mol with a single peak of 900 pmol/mol which is discussed in detail in the accompanying paper Fiedler et al. (2008). FLEXPART particle dispersion model analyses identified a region in North East China as the source region of the SO 2 emissions. 20

Initialization of the model
The air mass with the highest SO 2 mole fraction (around 10:40 UTC, e.g. in Fig. 3  Interactive Discussion has previously been described (e.g. Pirjola, 1999;Pirjola and Kulmala, 2001 (Fuchs and Sutugin, 1971) and aerosol coagulation (Fuchs, 1964). In this work a Lagrangian approach is used. From a prescribed SO 2 and OH 5 concentration the H 2 SO 4 concentration is calculated. Further model input needed is an initial particle concentration, pressure, relative humidity and temperature along the trajectory. The model then delivers the homogeneous binary nucleation rate J honu , condensation sink CS, which is principally the inverse H 2 SO 4 lifetime, and particle number size distributions between 0.86 nm and 1 µm. In this work we used 54 size bins.

10
The particle concentration calculation in the model moreover builds on a bimodal initial particle size distribution, possessing the lognormal parameters 15 and N 2 = 50 cm −3 (4) d 2 = 250 nm (5) with N 1 the initial particle number concentration in mode 1, d 1 the geometric mean 20 diameter of that mode and σ 1 the geometric standard deviation of the lognormal distribution, N 2 , d 2 and σ 2 respectively in mode 2. Here, four different scenarios (A, B, C, D) will be investigated for the highest SO 2 peak. Since no measurements existed for the preexisting particle size distribution, the initial particle concentrations are varied by multiplying the above values by 0, 1, 2 and 4, resulting in total initial particle concentrations of 0, 250, 500 and 1000 cm −3 . The simulation time is 8.5 days starting at 00:00 UTC on 25th of April 2006, so ending around noon on the 3rd of May 2006, which was the measurement flight day and time. The initial SO 2 concentration of 4.4×10 10 cm −3 was chosen in such a way that the final modeled SO 2 concentration (1.3×10 10 cm −3 , which corresponds at that temperature and pressure level to 900 pmol/mol) matches the measured SO 2 . At the beginning of 5 the simulation the air parcel under consideration was lifted to 8 km (≈335 hPa). Table 1 summarizes the model input data of the four model cases.
For the OH calculation a clear sky assumption was made. This is in accordance with a rough satellite cloud top temperature data analysis. Latitude, longitude and time in UTC have been taken into account as well as the length of the day. Typical maximum 10 OH concentrations were adopted from Logan et al. (1981). In Fig. 2 6 h mean values of temperature and relative humidity, RH, calculated from the water vapor concentration along the trajectory are plotted. The temperature stays rather constant around 240 K the whole time, relative humidity varies between 5% and 68% with two maxima at the beginning and on day 6 of the simulation. We also calculated the saturation vapor 15 pressure over ice based on the Magnus equation (Pruppacher and Klett, 2000) and the relative humidity over ice, RHice, also shown in Fig. 2. Since for temperatures below 0 • C the water vapor saturation pressure over liquid water is always higher than over ice, RHice is higher than RH for the same atmospheric water vapor concentration. However, water vapor supersaturation does not occur in the 8.5 simulation days. Interactive Discussion within the 8.5 days of simulation, SO 2 to decrease from 4.4×10 10 cm −3 (corresponding to a mole fraction of almost 3700 pmol/mol) to 1.3×10 10 cm −3 (corresponding to the measured maximum mole fraction of around 900 pmol/mol). In comparison, the FLEX-PART particle dispersion model predicted a SO 2 mole fraction of 2500-3500 pmol/mol at the location and time of our measurement (paper 1, Fig. 9, Fiedler et al., 2008).

5
Considering that FLEXPART treats SO 2 as an inert tracer, this compares well to the assumption of 3700 pmol/mol as initial SO 2 mole fraction input for the AEROFOR model. After 8.5 days, about 70% of the initially available SO 2 has undergone photochemical conversion to GSA, which in turn has been incorporated into aerosol particles. The H 2 SO 4 concentration (Fig. 3, upper panel) shows a diurnal variation, following the di-10 urnal variation of OH. Right in the beginning, H 2 SO 4 local noontime concentrations of 3×10 7 cm −3 are reached. The maximum varies slightly each day between 1×10 7 and 3×10 7 cm −3 . The homogeneous nucleation rate J honu (Fig. 3, lower panel) shows two strong maxima around noon on day 0 and day 1 (7000 and 20 000 cm −3 s −1 ) and three smaller peaks on day 2 and 6, which corresponds to the local minima in temperature 15 accompanied by local maxima in relative humidity. Low temperatures and high relative humidities favor new particle formation. The condensation sink CS (Fig. 4, lower panel) starts at 0 s −1 as no initial particles exist, but increases immediately to 0.01 s −1 simultaneously to the occurring nucleation. During nighttime the CS decreases caused by the decrease in the total particle surface, which results from coagulation and growth 20 of the existing particles. The surface to volume ratio decreases with increasing radius of the particles. So, if small particles coagulate to form bigger ones, the total aerosol surface will decrease. This means that the total aerosol surface available for gaseous sulfuric acid scavenging decreases during night, if no new particles are formed. After the new nucleation on simulation day 1 the CS reaches its maximum of 0.013 s −1 . 25 Eventually, this results in the particle concentrations depicted in Fig. 4, upper panel. A strong increase in the total particle concentration (up to 6×10 6 cm −3 ), caused by the two nucleation bursts on the first and second day of the simulation, is followed by a slow decrease of the total particles to a final value of ≈1600 cm −3 . Condensational and coagulational growth of the freshly formed particles forms particles with diameters larger than 30 nm already on the first day. Particles of these size classes may act as cloud condensation nuclei (CCN) and are therefore available for cloud formation. Similar graphs for model scenario B with an initial particle concentration of 250 cm −3 are not shown as no major changes occurred compared to scenario A. In model case 5 B, H 2 SO 4 concentration and homogeneous nucleation rate show nearly the same behavior as in model case A. The CS starts at 0.002 s −1 instead of 0 s −1 , because of the initial particles that are available for condensation already when the simulation starts. However, the CS maximum value is 0.013 s −1 as in scenario A. Freshly nucleated particles are formed in the same amount as in scenario A, so there is obviously enough 10 H 2 SO 4 available for both, growth of the initial particles and nucleation of new ones. The formation of CCN after 1 day is still enhanced and a final total particle concentration of 1300 cm −3 at the measurement site can be expected.
The plots for model scenario C (initial particle concentration 500 cm −3 ) are shown in Figs. 5 and 6. The higher initial particle concentration again has no substantial influ-15 ence on the H 2 SO 4 concentration and the nucleation rate, but the higher CS (starting at 0.005 s −1 and 0.0135 s −1 in maximum) slows down the growth of the particles particularly in the size classes N50, N100 and N200. The growth starts one day later than in scenario A and the total increase in the number concentration of CCN is less developed. Nevertheless the final concentration of CCN reaches in the sum of initial and 20 freshly formed particles 1200 cm −3 .
Model scenario D, depicted in Figs. 7 and 8, eventually starts with the assumption of 1000 cm −3 as initial particle concentration, which is almost the final value in scenario C. The H 2 SO 4 concentration development stays nearly the same, but the nucleation rate J honu is slightly lowered. The condensation sink is further increasing and is all the 25 time above 0.01 s −1 with a maximum of 0.0143 s −1 . The increase in the total particle concentration on the first two simulation days is still appreciable, but the growth in all size classes is now markedly reduced. So the H 2 SO 4 concentration seems still to be high enough for new particle formation, but a large amount of H 2 SO 4 will be Typical previously measured particle concentrations for background aerosol are around 500 cm −3 and in very polluted cases also several 1000 cm −3 can be reached 5 (e.g., Minikin et al., 2003). So at least model scenario C seems to be quite realistic.
With even higher initial particle concentrations (e.g. 2000 cm −3 ) nucleation would not occur at all due to a complete removal of the condensable gases by condensation onto the preexisting aerosol. Figure 9 gives an overview of the modeled final particle concentrations N30, N100 10 and N200 (after 8.5 days). For comparison, the corresponding initial particle concentrations are also given. The right panel of Fig. 9 shows model predictions with H 2 SO 4 formation switched off. Here new aerosol formation does not occur and preexisting particles grow only by coagulation. For increasing preexisting particle concentrations coagulation becomes more efficient resulting in a decrease of N30 and N100 with re-15 spect to the corresponding concentrations at time 0. In contrast, N200 increases but the increase is only very small. When photochemical SO 2 conversion to H 2 SO 4 is switched on, the situation changes drastically (Fig. 9, left panel). Now, for each of the 4 scenarios A, B, C and D, the final concentrations of N30, N100 and N200 are much larger. Hence, the AEROFOR model simulations indicate very substantial SO 2 -20 mediated growth of preexisting and newly formed aerosol particles.

Sensitivity analyses
The simulation of the model case C was repeated to test the sensitivity of the concentration of CCN after 8.5 days against the initial SO 2 concentration, OH concentration, nucleation rate and initial particle size distribution. 25 When multiplying the nucleation rate by a factor of 100 and keeping the OH and SO 2 concentrations as in case C the final N50, N100 and N200 increased only 4%, 7% and 2%, respectively. The total particle concentration became threefold but the major part 2771 Introduction

Tables Figures
Back Close

Printer-friendly Version
Interactive Discussion of the particles remained smaller than 4 nm in diameter size. Thus we can conclude that the CCN production is much more sensitive to particle growth, i.e. condensable vapor concentration, than to the actual nucleation rate. Since the GSA formation rate is proportional to OH×SO 2 , the simulations were repeated by multiplying first the OH by a factor of 2 (SO 2 as in case C) and then the initial 5 SO 2 by a factor of 2 (OH as in case C). The final SO 2 concentration was 4.5×10 9 cm −3 (≈305 pmol/mol) and 2.8×10 10 cm −3 (≈1900 pmol/mol), respectively. The increases in N50, N100 and N200 were 16%, 41% and 35%, and 25%, 67% and 92%, respectively.
The other factor which affects the growth of nucleated particles is the condensation sink CS of the preexisting particles. The CS is mainly determined by the num-10 ber concentration and the size distribution of the preexistent particles. Therefore, a CS sensitivity study is already included in the 4 simulation cases. The CS of case B (initial particle concentration of 250 cm −3 ) e.g. corresponds to the CS of case C (initial particle concentration of 500 cm −3 ) if the diameter of the particles in mode 1 of case C (N1=400 cm −3 ) is changed to 62 nm and the diameter in mode 2 of case C 15 (N2=100 cm −3 ) is changed to 124 nm (see also Table 1 for comparison). On the other hand, if the diameters of the particles in mode 1 and 2 of case D are changed to 100 nm and 200 nm, respectively, the CS of case B is the same CS as in case D.

Implications for CCN
Our above findings have interesting implications for CCN and cloud droplet forma-20 tion. In the atmosphere, water vapor supersaturation (WSS) occasionally occurs and aerosol particles with sufficiently large diameters and hygroscopicity may act as CCN and induce water cloud droplet formation. Mostly, WSS is induced by adiabatic cooling or mixing of air masses with different temperatures and humidities (e.g. Seinfeld and Pandis, 2006;Gettelman et al., 2006). Typical atmospheric water vapor supersaturation rarely surpasses 2%, the median of observed supersaturation even is 0.1% (Pruppacher and Klett, 2000 Sulfuric acid is very hygroscopic and therefore H 2 SO 4 represents an ideal CCN material. H 2 SO 4 -H 2 O aerosols are liquid in the temperature and RH conditions encountered in the troposphere. While H 2 SO 4 contained in an aerosol droplet may promote water vapor uptake, it may at least initially hinder droplet freezing depending on the H 2 SO 4 fraction of the droplet mass (Carleton et al., 1997;Ettner et al., 2004). As 5 water vapor uptake proceeds and the droplet size has increased sufficiently and the H 2 SO 4 mass fraction has decreased sufficiently, the droplet may ultimately freeze even quasi-homogeneously, depending on RHi. However, a pure water droplet only freezes at Temperatures <235 K and RHi>145%. Those values have not been reached during the 8.5 simulation days (Fig. 2). Figure 10 depicts the minimum diameter a H 2 SO 4 -H 2 O 10 aerosol particle needs to possess in order to become a water vapor condensation nucleus (activation diameter). For example for WSS=0.5%, the critical diameter is about 56 nm. As suggested by our above AEROFOR model simulations, in the Chinese pollution plume even newly formed particles grow to diameters larger than 50 nm within only a few days and therefore will act as CCN at WSS=0.5%. For lower WSS the minimum 15 diameter increases very steeply (Fig. 10), which is due to the Kelvin effect. Figure 11 shows the modeled number concentration of CCN sized H 2 SO 4 -H 2 O aerosol particles versus WSS. CCN concentrations refer to the end of the simulation period and to the four model cases A (Nini=0), B (Nini=250) etc. The right panel is the model with gaseous sulfuric acid formation switched off. Here particles grow only due 20 to coagulation. In this case for models B, C and D the CCN concentrations increase initially steeply with WSS which means that an increasing fraction of the particles can act as CCN. For WSS>0.4% (critical diameter: 65 nm), CCN concentrations remain nearly constant, which means that all particles present at the end of the simulation period act as CCN. Due to coagulation, during the simulation period, their total num-25 ber concentration has decreased and their diameters have increased. The left panel of Fig. 11  Our findings suggest that, already on day 6, the Chinese pollution plume had developed its full CCN potential (maximum N100) and was primed for the formation of large concentrations of relatively small water droplets whenever WSS of about 0.2% 10 built up (activation diameter 100 nm for 0.2% WSS, view Fig. 10). Condensation of a given mass of supersaturated water vapor on more CCN leads to smaller and more numerous cloud droplets. This tends to increase the albedo of the cloud and to increase the cloud lifetime with respect to rainout, since the droplet sedimentation velocity is approximately proportional to the square of the droplet diameter. In other words, after 15 9 days and at a distance of about 20 000 km from its birthplace, the plume had very suitable conditions for the formation of whiter clouds possibly possessing a reduced tendency for rainout. The Chinese pollution plume reported here represents a striking example of an environmental impact of long range transport of fossil fuel combustion generated SO 2 . 20