Air pollution control and decreasing new particle formation lead to strong climate warming

The number concentration of cloud droplets determines several climatically relevant cloud properties. A major cause for the high uncertainty in the indirect aerosol forcing is the availability of cloud condensation nuclei (CCN), which in turn is highly sensitive to atmospheric new particle formation. Here we present the effect of new particle formation on anthropogenic aerosol forcing in present-day (year 2000) and future (year 2100) conditions. The present-day total aerosol forcing is increased from−1.0 W m−2 to −1.6 W m−2 when nucleation is introduced into the model. Nucleation doubles the change in aerosol forcing between years 2000 and 2100, from +0.6 W m−2 to +1.4 W m−2. Two climate feedbacks are studied, resulting in additional negative forcings of −0.1 W m−2 (+10 % DMS emissions in year 2100) and −0.5 W m−2 (+50 % BVOC emissions in year 2100). With the total aerosol forcing diminishing in response to air pollution control measures taking effect, warming from increased greenhouse gas concentrations can potentially increase at a very rapid rate.

Sulphuric acid is thought to be the main precursor for atmospheric nucleation (Kulmala et al., 2006;Kerminen et al., 2010;Sipilä et al., 2010), hence the emissions of oceanic dimethyl sulphide (DMS) and both natural and anthropogenic sulphur dioxide are expected to dominate the new particle formation process.Anthropogenic SO 2 emissions have increased by three orders of magnitude in the last 250 yr, from 0.06 Tg(S) yr −1 in the year 1750 to 54 Tg(S) yr −1 in the year 2000 (Dentener et al., 2006).However, the global SO 2 emissions might already have peaked in the early 1970s (Smith et al, 2011), and the application of flue-gas desulfurization in power plants has decreased emissions in China by 9.2 % between 2006 and 2010 (Lu et al., 2011).Due to the relatively easy suppression of SO 2 emissions, the projected future emission pathways show a strong decrease in global emissions.Recently, "Representative Concentration Pathways" (RCPs) (Moss et al., 2010;Lamarque et al., 2011) have been introduced for the scenario development process of IPCC AR5.All RCPs project a significant decrease in global anthropogenic SO 2 emissions, with annual emissions ranging from 6.5 Tg(S) to 13 Tg(S) in the year 2100.Not only the SO 2 emissions are expected to decrease: global emissions of fossil fuel organic and black carbon are predicted to peak before 2030 and then to decline rapidly, such that the levels at 2100 would be close to those at 1850.However, most RCPs predict a continuous increase of ammonia emissions, at 2100 being 3 to 5 times those at 1850.Future emissions of biomass burning aerosols behave very differently between the different RCPs in terms of both the sign of the trend and absolute emission levels.The estimated present-day to future changes in oceanic DMS emissions are rather small: +6.5 % (Ramaswamy et al., 2001), +2.4 % (Bopp et al., 2004) and −8.0 % (Stier et al., 2006).
Despite its importance for nucleation, sulphuric acid can explain only a fraction of the subsequent nuclei growth.Oxidation products of biogenic volatile organic compounds (BVOCs) may provide the required growth for aerosols to reach CCN sizes (Jimenez et al., 2009;Paasonen et al., 2010).In addition, certain organic compounds might also play a role in the first steps of nucleation (Metzger et al., 2010).However, the centennial-scale changes in BVOC emissions are highly uncertain, ranging from almost zero to increases by more than 70 %, depending on which processes are considered in the underlying emission algorithms and prognostic vegetation models (Lathière et al., 2005;Arneth et al., 2007).
Our aim is to study aerosol-cloud-climate interactions in three emission environments: pre-industrial, present-day and future.We provide a quantification of the impact of aerosol nucleation on particle and cloud droplet number concentrations and anthropogenic aerosol forcing, accounting also for the possible impact of increase in biogenic precursor emissions (BVOCs and oceanic DMS).

Global aerosol-climate model ECHAM5-HAM
We use the global climate model ECHAM5-HAM (Stier et al., 2005) to simulate aerosol concentrations, cloud properties and total aerosol forcing with pre-industrial (year 1750), present-day (year 2000) and future (year 2100) emissions.The model is run in a T42 spectral resolution, corresponding to a mean horizontal resolution of 2.8 • .In the vertical, 31 levels extend from surface up to 10 hPa.ECHAM5-HAM includes the aerosol microphysics model M7 (Vignati et al., 2004), which considers dust, sea salt, black carbon (BC), particulate organic matter (OC) and sulfate.Nitrate aerosols are not included in the model.Aerosol size distribution is described by four soluble (internally mixed) and three insoluble (externally mixed) log-normal modes: nucleation-mode (particle radius r < 5 nm), Aitken-mode (5 nm < r < 50 nm), accumulation-mode (50 nm < r < 500 nm) and coarse-mode (r > 500 nm).The mode standard deviations are constant: 2.00 for the coarse modes and 1.59 for other modes.Binary sulphuric acid nucleation (Vehkamäki et al., 2002) is included throughout the atmosphere, and activation-type nucleation (Kulmala et al., 2006) is constrained to the boundary layer.We assume monoterpene oxidation products to form SOA with a 15 % yield, and this organic mass is partitioned to the aerosol phase according to condensation sink of each mode (Makkonen et al., 2009) in the boundary layer during one timestep.No organics are traced in the gas-phase.The modal setup of M7 is not ideal for studies of new particle formation: for example, nucleation mode can not keep track of newly formed particles and at the same time model the growth of aged particles.To improve this, we convert the nucleated particles to 3 nm size by using a parameterization by Kerminen and Kulmala (2002).Some of the numerical problems in the modal approach could be avoided by using sectional models, for example SALSA (Kokkola et al., 2008;Bergman et al., 2011).
The aerosol model is coupled to a double-moment cloud microphysics scheme (Lohmann et al., 2007), and cloud droplet activation is parameterized according to Lin and Leaitch (1997).The number concentration of particles with wet radii larger than 35 nm, N a is given to the parameterization, which calculates the number of activated cloud droplets from N a , updraft velocity and cloud droplet concentration from the previous timestep.The updraft velocity is obtained from the grid-mean vertical velocity, turbulent kinetic energy and convectively available potential energy.The activation radii of 35 nm lies in the middle of the Aitken mode, however it is assumed that only the larger part of the log-normal mode is able to activate.Only particles in the Aitken, accumulation and coarse modes can activate into cloud droplets.The model assumes a minimum cloud droplet number concentration of 40 cm −3 , which was the highest value found in a model intercomparison by Quaas et al. (2009).There is evidence that this unphysical constraint affects the aerosol indirect effect (Hoose et al., 2009), which might result in underestimation of total aerosol forcing in this study.
ECHAM5-HAM utilises monthly-mean oxidation fields of OH, H 2 O 2 , O 3 and NO 2 , which are prescribed based on simulations with the MOZART chemical transport model (Horowitz et al., 2003).SO 2 is oxidized by OH in the gas phase and by H 2 O 2 and O 3 in the aqueous phase.The gas phase reactions of DMS with OH and NO 3 are considered.The prescribed oxidant fields represent conditions of the year 2000, but we apply them also for the pre-industrial and future simulations.
The sea-surface temperature and sea ice extent are prescribed to present-day values.We integrate the model for 5 yr and calculate the total aerosol forcing as radiative flux perturbation (Haywood et al., 2009;Lohmann et al., 2009).The flux perturbation is calculated for total radiation (short-wave+longwave).The technique is similar to what most models used in the indirect effect intercomparison study by Quaas et al. (2009).Since our model setup couples aerosols with cloud microphysics, the total effect includes the direct effect, semi-direct effect, 1st indirect effect (cloud albedo), and indirect effects beyond the cloud albedo effect: although the prescribed sea-surface temperatures will keep the overall climate fixed, changes in aerosols and cloud properties can induce changes in e.g.precipitation and cloud cover.
We will conduct two sets of experiments, with and without nucleation.In simulations without nucleation both boundary layer nucleation and binary sulphuric acid-water nucleation is turned off, but other microphysical processes are left untouched (e.g.sulphuric acid and BVOC oxidation products can still condense on existing particles).

Anthropogenic emissions
Anthropogenic aerosol and precursor emissions are taken from the AeroCom inventories for the years 1750 and 2000 (Dentener et al., 2006).For the future we apply three of the four pathways constructed for the IPCC AR5 emission development process, namely RCP 3-PD (van Vuuren et al., 2007), RCP 4.5 (Smith and Wigley, 2006;Clarke et al., 2007;Wise et al., 2009) and RCP 8.5 (Riahi et al., 2007).The evolution of aerosols in these pathways has been presented in detail by Lamarque et al. (2011), who found a strong decrease in anthropogenic aerosols by the year 2100.The anthropogenic emissions of aerosols and precursors are projected to be greatly reduced by the year 2100.The emissions of SO 2 will decrease from 107 Tg(SO 2 ) yr −1 to 13-26 Tg(SO 2 ) yr −1 .The emissions of BC will decrease from 7.8 Tg yr −1 to 3.4-4.4Tg yr −1 , showing a narrower range of values in the year 2100 than SO 2 .The emissions of OC will decrease from 36 Tg yr −1 to 19-32 Tg yr −1 .The only aerosol precursor showing a future increase is NH 3 , which will increase from present-day emission of 49 Tg yr −1 to 53-82 Tg yr −1 due to agricultural emissions.
The shapes of number size distributions of emitted primary particles are assumed identical in all simulations, following Stier et al. (2005).The primary emissions of OC and BC from fossil-fuel and bio-fuel are assumed to have a number mean radius of 30 nm.Wildfire emissions are assumed to lead to larger primary OC and BC particles with the number mean radius of 75 nm.The primary emission radii are based on AeroCom recommendations (Dentener et al., 2006), but have been adapted to the log-normal setup of M7 (Stier et al., 2005).As a result, the used primary aerosol radii are larger than those suggested by AeroCom, resulting in fewer emitted primary particles.Although this has some effects on simulated number concentrations, it is not straightforward to test the sensitivity to the emission radii.Certain radii suggested by AeroCom (Dentener et al., 2006) fall in between the modes of M7, which would create undesired numerical effects in aerosol microphysics.
Due to the rapid gas-to-particle conversion near source areas, we followed Stier et al. (2005), assuming that 2.5 % of all emitted SO 2 (except from DMS) is converted to primary sulfate and partitioned to the Aitken, accumulation and coarse modes.None of these particles are included in the nucleation mode, even though Stevens et al. (2012) highlighted the importance of nucleation in power plant plumes.

Natural emissions
Natural emissions of sea salt and dust are modeled interactively in all simulations (Stier et al., 2005), hence their emission strength might differ between experiments e.g. in response to changes in the wind speed.The emissions of volcanic sulphur are identical in all the simulations performed corresponding to present-day emissions (Stier et al., 2005).
The oceanic DMS emissions are modeled interactively according to Kettle and Andreae (2000) and Nightingale et al. (2000), which results in annual global DMS emission of 23.4 Tg(S) (Stier et al., 2005).In order to test the DMS emission sensitivity part of CLAW hypothesis (Charlson et al., 1987), we performed a future scenario simulation where oceanic DMS emissions are increased by 10 %.The applied increase is higher than proposed in literature (+2.4 % in Ramaswamy et al., 2001, +6.5 % in Bopp et al., 2004).
For biogenic VOC emissions we consider only monoterpene emissions, which leads to an underestimation of SOA.The monoterpene emissions are prescribed monthly fields generated with the dynamic vegetation model LPJ-GUESS (Schurgers et al., 2009).Depending on whether the increasing atmospheric CO 2 concentration affects monoterpene production in the same way as it does for isoprene (Schurgers et al., 2009), these emissions may remain more or less unchanged, or they increase in response to warmer climate and higher vegetation productivity.The annual global monoterpene emissions are 31, 30 and 28 Tg(C) for pre-industrial, present-day and future, respectively.The applied monoterpene emissions are on the lower side of values reported in the literature, ranging from 30 to 128 Tg(C) yr −1 (Carslaw et al., 2010).To assess the responsiveness of the vegetation to increase in temperature, we performed an additional future scenario where future BVOC concentrations are increased by 50 % from the base-level of 28 Tg(C).aerosol number concentration only in regions with intensive wild-land fires (Africa, South America, North-East Eurasia).

Evolution of number concentrations
The only location showing a decrease in CN concentration since the pre-industrial period is high-latitude boreal forests in North America and eastern Siberia, where increased fire suppression has decreased emissions (Brenkert et al., 1997).
There is a clear change in the spatial distribution of primary emissions when moving to the year 2100: in general, number concentrations increase in India and Africa, and decrease elsewhere.
When aerosol nucleation is turned on (Fig. 1, lower panels), CN concentrations are increased significantly.Preindustrial anthropogenic sources of SO 2 were low, hence aerosol nucleation originated mostly from natural precursors.Some of these sources can be identified in lower panels of Fig. 1: volcanoes (Philippines, areas west of New Caledonia, and southern Japan) and DMS (the band around 60 • S).The human impact is clearly visible under present-day conditions.Increases in primary emissions and new particle formation lead to doubled CN concentrations in US, Europe, Middle East, India, China and South Africa.There are several large areas inside these regions where CN concentrations are up to five times above the pre-industrial values.There are significant reductions in CN concentrations from present-day to the year 2100: aerosol number concentrations are decreased by 50-90 % in North America, Europe, Eastern China and South Africa.Certain equatorial regions, such as Middle Africa, show increased CN concentrations due to increases in wildfires.
The present-day CN concentrations simulated in this study are similar to those in Fig. 7e by Yu et al. (2010), who also considered activation-type nucleation in the boundary layer.
However, the land-ocean contrast in the CN concentration is stronger in our study.The global average CN concentration of 3017 cm −3 simulated here exceeds the estimate of 1063 cm −3 by Merikanto et al. (2009), but comparing the spatial CN concentration distribution modeled here to Fig. 5 in Merikanto et al. (2009) indicates that a major part of the discrepancy might be due to nucleation in the DMS belt around 60 • S: Merikanto et al. (2009) shows CN concentrations of <1000 cm −3 while our simulations indicate concentrations of about 3000 cm −3 .Over the continents, our results seem to be relatively constistent with those by Merikanto et al. (2009).
Figure 2 presents values for pre-industrial, present-day and future CCN (cloud condensation nuclei, d p > 70 nm) concentrations.Due to extremely low pre-industrial CCN concentrations without nucleation (upper panels), the presentday anthropogenic emissions introduce a global average increase by a factor of 4 to the boundary layer CCN concentration.Similarly to CN concentrations without nucleation, CCN concentrations increase in Africa and India by year 2100, but otherwise show a decrease.
When nucleation is turned on (lower panels), the relative present-day anthropogenic effect on CCN concentrations is smaller.The present-day emissions lead to doubled CCN concentrations in almost all locations above land, as compared with the pre-industrial values.CCN concentrations increase more than fivefold in North America, central Europe, India and China, compared to pre-industrial.The presentday CCN distribution compares well with that by Spracklen et al. (2008), although our study finds slightly higher CCN concentration over the oceans.Even with decreasing global emissions, CCN concentrations in the year 2100 are still doubled to the pre-industrial values in India, Middle East, and some parts of Europe and Africa.The CCN hotspots are similar in simulations with and without nucleation.The simulated overall strong future decrease in the availability of condensation nuclei for cloud droplets are expected to lead to drastic changes in cloud properties.
The global average concentrations of CN, CCN and cloud droplet number are summarized in Table 1.In a preindustrial world without nucleation, the total aerosol number concentration of 22 cm −3 was mainly of CCN size.The particle number concentrations were mostly below 40 cm −3 , which is the applied minimum limit for cloud droplet number concentration.Anthropogenic primary emissions lead to a factor of 5 increase in CN concentration by the year 2000.Future CN concentrations from primary emissions either decrease slightly (RCP 3-PD, RCP 4.5) or increase due to increased emissions in Africa and India (RCP 8.5).Turning on nucleation in the model increases the CN concentration by a factor of 100, 25 and 20 in the pre-industrial, present-day and future conditions, respectively.Without nucleation, the relative increase in the CN concentration from pre-industrial to future is 500 %.Including nucleation increases the preindustrial CN concentration significantly, causing relative increase of only 5-9 % between pre-industrial and future.
The global 5-yr-average CDNC at the cloud-top height with and without nucleation are shown in Fig. 3. Including nucleation in the model increases the overall global CDNC by a factor of 2-3 compared with CDNC without nucleation.Our model shows a 70 % increase in the boundary layer CCN concentration (from 236 cm −3 in year 1750 to 397 cm −3 in year 2000) and 20 % increase in the cloudtop CDNC from the year 1750 to 2000 (from 167 cm −3 to 203 cm −3 ).Wang and Penner (2009) found a rather similar anthropogenic perturbation of +85 % to boundary layer CCN.Pierce and Adams (2009)  of +220 % and +200 % in the tropospheric CCN(0.2 %) concentration with binary and ternary nucleation, respectively.The CDNC in the altitude range 300-1000 m was found to increase from 145 cm −3 (year 1850) to 240 cm −3 (year 1750) by Merikanto et al. (2010), which is much more than the CDNC increase found in our study (but identical to the 70 % increase we found for the CCN concentration).Spracklen et al. (2008) found a 10 % increase in the CCN(0.2%) concentration and 20 % increase in CCN(1.0 %) concentration due boundary layer nucleation when using the same activation coefficient.Although our results presented here compare simulations with nucleation turned off completely, our model shows an increase of about 20-30 % in the boundary layer CCN due to activation-type nucleation.The sensitivity of our modeled CCN and CDNC to present-day anthropogenic perturbation seems to be consistent with earlier literature, although our result is on the smaller side.This might be due to overestimation of the pre-industrial CCN and CDNC due to nucleation.Also the minimum CDNC of 40 cm −3 affects the pre-industrial CDNC, and the relative changes in CDNC. Figure 3 also shows that the reductions in primary aerosol and aerosol precursor emissions by the year 2100 lead to a strong decrease in CDNC.In simulations without nucleation, the decrease is 17 cm −3 (22 %).When nucleation is turned on, the relative decrease is smaller (16 %), which would indicate that nucleation actually slows the decrease in cloud albedo.The CDNC in the year 2100 ranges from 167 cm −3 to 172 cm −3 (red shading in Fig. 3), showing very little variation due to anthropogenic emission uncertainties.It should be noted that without nucleation, the CDNC in the year 2100 is still 15 % higher than in year 1750, while the future CDNC Additional simulations with +50 % biogenic VOC emissions (green) and +10 % oceanic DMS emissions (blue) are simulated with the anthropogenic emission projection RCP 4.5.The markers represent average forcings for present-day (♦), RCP 3-PD (×), RCP 4.5 ( ) and RCP 8.5 ( ).The errorbars correspond to the range of one standard deviation, calculated from variation during 5-yr simulation.
in nucleation simulations it is only 2 % higher than in preindustrial period.
The climate feedback simulations show a clear increase in CDNC, +2 % and +5 % for an emission increase of +10 % DMS and +50 % BVOC, respectively.The two experiments work in very different ways: the increase in DMS emission leads to increase in both nucleation and aerosol growth over oceans, whereas the increase in BVOC emission only provides more growth for particles over land.

Total aerosol forcing
Figure 4 shows the total anthropogenic aerosol forcing for the years 2000 and 2100.The results clearly indicate how accounting for atmospheric nucleation is critical for aerosol forcing calculations.In the present-day atmosphere, including new particle formation increases the magnitude of the total aerosol forcing from −1.03 W m −2 to −1.61 W m −2 , indicating more cooling from anthropogenic aerosols.Our present-day forcing estimates are within the 90 % confidence range by (Forster et al., 2007), from −2.2 to −0.5 W m −2 , which does not consider the cloud lifetime effect included here.Hansen et al. (2011) used observations of the ocean heat uptake and global temperature change, and found the aerosol climate forcing at 2010 to be −1.6 ± 0.3 W m −2 , which is identical to our study.The total aerosol forcing of −1.61 W m −2 simulated here is also similar to −1.57 ± 0.66 W m −2 obtained for the total aerosol forcing averaged over ten global models by Quaas et al. (2009).While nucleation increases the present-day aerosol forcing, more importantly the change in aerosol forcing from present-day to year 2100 is +1.3-1.4W m −2 when nucleation is considered, while it is only +0.5-0.8W m −2 without nucleation.It should be noted that the minimum CDNC of 40 cm −3 applied here can be considered rather high especially for preindustrial clouds, hence causing an underestimation of the anthropogenic forcing (Hoose et al., 2009).
All simulations performed in this study use coupled aerosol-cloud microphysics and prescribed sea-surface temperatures.Calculation of aerosol direct forcing would require fixing of meteorology, i.e. nudging the model.To give some idea on the contribution of direct/indirect effects on the simulated total aerosol effect, we diagnosed the clear-sky top-ofatmosphere shortwave aerosol direct forcing.We can compare these forcings to Lamarque et al. (2011), who used the same emission pathways and calculated the clear-sky top-ofatmosphere shortwave aerosol forcing.We can also compare our present-day forcing to AeroCom intercomparison study by Schulz et al. (2006).Both our present-day simulations show a −0.70 W m −2 clear-sky forcing, which is almost identical to −0.68 ± 0.28 W m −2 in Schulz et al. (2006) and slightly smaller than −0.8 W m −2 in Lamarque et al. (2011).In simulations without nucleation, the direct aerosol forcing is decreased to −0.26 W m −2 (RCP 3-PD), −0.22 W m −2 (RCP 4.5) and −0.28 W m −2 (RCP 8.5) by the year 2100.These findings are similar to Lamarque et al. (2011), who found a future aerosol direct forcings of −0.24 W m −2 (RCP 3-PD), −0.24 W m −2 (RCP 4.5) and −0.37 W m −2 (RCP 8.5).The slight discrepancy is partly explained by the future increase in ammonium nitrate forcing by 0-0.05 W m −2 , which is not considered in our study.Including nucleation in the model decreases the future clear-sky aerosol direct forcing to −0.05 W m −2 .In the AeroCom intercomparison (Schulz et al., 2006), the present-day clear-sky direct forcing of −0.68 ± 0.28 resulted in an all-sky direct forcing of −0.22 W m −2 , hence it is very likely that a major part of the total aerosol forcing shown in this study is due to aerosol indirect effects.
The applied emission pathways show strong changes in anthropogenic emissions of black carbon (BC, 40-60 % reduction) and organic carbon (OC, 10-60 % reduction).It has been shown that a 50 % reduction in the primary BC/OC mass and number emissions alone would lead to +0.19 W m −2 forcing in present-day climate (Chen et al., 2010), when taking into account both the cooling from reduced BC (−0.12 W m −2 ) and simultaneous warming from decreased CDNC (+0.31 W m −2 ).The findings by Lamarque et al. (2011) indicate a 0.06-0.09W m −2 positive clear-sky forcing from reduced BC, when applying the same emission pathways as in this study.Hence, the BC/OC reduction could explain a major part of the simulated future warming without nucleation (+0.64 W m −2 ).Kloster et al. (2008) showed that even a +1.13 W m −2 warming until 2030 from the reduced aerosol load would be plausible under a "maximum feasible reduction" sce-nario.Although their present-day total aerosol forcing (−2.00 W m −2 ) is stronger compared with our simulations, the applied 80 % reduction in anthropogenic SO 2 is comparable to future scenarios used here.
While the simulated present-day total aerosol forcing (−1.61 W m −2 ) is rather close (but of opposite signal) to the present-day forcing of CO 2 (+1.66 W m −2 , Forster et al., 2007), the future warming from applied air pollution control and the resulting decrease in aerosol effects (a change of +1.38 W m −2 in total aerosol forcing, when averaged over future scenarios) would accelerate the rate of climate warming in response to increasing greenhouse gas concentrations.For example, in the scenario of RCP 3-PD (van Vuuren et al., 2007) the CO 2 forcing is estimated to increase from +1.5 W m −2 (year 2000) to +2.3 W m −2 (year 2100), hence the warming resulting from reduced aerosol forcing alone might be more than that due to CO 2 concentration over the same time period.
Could changes in natural aerosol precursors counteract the effects of anthropogenic emission changes?This seems unlikely, as shown in Fig. 4. The simulation with +50 % BVOC emission shows increased particle growth from oxidized organics, leading to 5 % increase in CDNC and finally an additional negative forcing of −0.46 W m −2 .Similarly, an increase of +10 % in DMS emission can only add an extra cooling of −0.12 W m −2 to the future aerosol forcing.It should be noted, however, that the effect of these additional emissions is not negligible, and that there might be other processes and feedbacks involved.

Conclusions
Our analysis provides evidence on how the present-day climate forcing by aerosol particles is controlled strongly by atmospheric new particle formation.This is due to a favourable combination of anthropogenic sulphur and biogenic organic vapour emissions, which together produce effectively new cloud condensation nuclei that enhance the negative cloud forcing.
We have shown that even though the total number concentration of primary aerosols might stay constant or even increase slightly by the year 2100, the strong reduction in anthropogenic SO 2 emissions will lead to less aerosol nucleation, rendering the total number concentrations close to pre-industrial values.More importantly, due to the simulated strong coupling of SO 2 and CCN, the future cloud droplet concentrations will be decreased to almost pre-industrial concentrations.Hence, the current cooling effect by aerosol particles (total aerosol forcing of −1.61 W m −2 ) will be reduced close to zero (−0.23 W m −2 ) by 2100.The two climate feedbacks studied (increase in either DMS or BVOC) show increased CDNC and result in additional negative forcing, yet they can not compensate for the effects of decreasing anthropogenic emissions.The suppression of future aerosol forcing, if true, would lead to even a larger warming of climate than what is expected based on increased greenhouse gas concentrations (Andreae et al., 2005).

Fig. 1 .
Fig.1.Annual average concentration (cm −3 ) of CN (condensation nuclei, aerosol particles with diameter d p > 3 nm) in the boundary layer for the years 1750 (left), 2000 (center) and 2100 (right).The results for the year 2100 are an average from all three future emission pathways (RCP 3-PD, RCP 4.5, RCP 8.5).The upper panels show results from simulations without nucleation, and lower panels with nucleation included.

Fig. 2 .
Fig. 2. Annual average concentration (cm −3 ) of CCN (cloud condensation nuclei, d p > 70 nm) in the boundary layer for the years 1750 (left), 2000 (center) and 2100 (right).The results for the year 2100 are an average from all three future emission pathways (RCP 3-PD, RCP 4.5, RCP 8.5).The upper panels show results from simulations without nucleation, and lower panels with nucleation included.

Fig. 3 .
Fig. 3.Global 5-yr average cloud droplet number concentration (cm −3 ) for the pre-industrial, present-day and future conditions.The cloud droplet number concentration is sampled at the cloud-top height from simulations with (red) and without (black) nucleation.Error bars and corresponding shading indicate the range in CDNC due to different future anthropogenic emission pathways.Additional simulations with +50 % biogenic VOC emissions (green) and +10 % oceanic DMS emissions (blue) are simulated with the anthropogenic emission projection RCP 4.5.

Table 1 .
Annual global average boundary layer CN and CCN concentration (cm −3 ), and cloud-top CDNC (cm −3 ) in simulations with and without nucleation.