The oxidation of biogenic volatile organic compounds (BVOCs) gives a range of products, from semi-volatile to extremely low-volatility compounds. To treat the interaction of these secondary organic vapours with the particle phase, global aerosol microphysics models generally use either a thermodynamic partitioning approach (assuming instant equilibrium between semi-volatile oxidation products and the particle phase) or a kinetic approach (accounting for the size dependence of condensation). We show that model treatment of the partitioning of biogenic organic vapours into the particle phase, and consequent distribution of material across the size distribution, controls the magnitude of the first aerosol indirect effect (AIE) due to biogenic secondary organic aerosol (SOA). With a kinetic partitioning approach, SOA is distributed according to the existing condensation sink, enhancing the growth of the smallest particles, i.e. those in the nucleation mode. This process tends to increase cloud droplet number concentrations in the presence of biogenic SOA. By contrast, an approach that distributes SOA according to pre-existing organic mass restricts the growth of the smallest particles, limiting the number that are able to form cloud droplets. With an organically mediated new particle formation mechanism, applying a mass-based rather than a kinetic approach to partitioning reduces our calculated global mean AIE due to biogenic SOA by 24 %. Our results suggest that the mechanisms driving organic partitioning need to be fully understood in order to accurately describe the climatic effects of SOA.
Biogenic volatile organic compounds (BVOCs), such as monoterpenes and isoprene, are emitted into the atmosphere by vegetation (Guenther et al., 1995, 2006) and are rapidly oxidised. The oxidation of BVOCs yields products with lower volatility, which may partition to the particle phase forming secondary organic aerosol (SOA). Organic compounds contribute a large fraction of submicron aerosol mass (Murphy et al., 2006; Zhang et al., 2007; Jimenez et al., 2009) with important impacts on air quality and climate (Fiore et al., 2012; Scott et al., 2014).
The part of the aerosol size distribution to which SOA is added affects the number, size and composition of particles in the atmosphere; in particular, the number of particles that are able to act as cloud condensation nuclei (CCN). The availability of CCN controls cloud droplet number concentrations (CDNC) and subsequently cloud albedo; therefore, the manner in which organics are distributed has potential implications for the first aerosol indirect effect (AIE) of biogenic SOA.
The presence of SOA can affect atmospheric CCN concentrations in several ways. Firstly, the condensation of SOA may grow particles to larger sizes, increasing CCN concentrations (Riipinen et al., 2012). However, this enhanced growth increases the condensation sink for potential nucleating vapours and the coagulation sink for nucleation mode particles. The net change to CCN concentration therefore reflects the competition between particle growth and the scavenging of particles and vapours. Secondly, condensation of water-soluble organic species can make hydrophobic particles more hydrophilic, providing an additional source of new CCN (Petters et al., 2006).
The transfer of semi-volatile gas-phase organic species into the condensed phase is often treated assuming thermodynamic equilibrium (Pankow, 1994; Odum et al., 1996). When simulating the evolution of the aerosol size distribution, a consequence of assuming instant equilibrium is that the net condensation of new organic mass scales with the existing organic aerosol mass size distribution (Kroll and Seinfeld, 2008; Pierce et al., 2011; Riipinen et al., 2011). Because aerosol mass scales with diameter cubed, small particles that require condensational growth to reach climatically relevant sizes receive only a trivial fraction of the new SOA and subsequently do not grow.
However, if the volatility of organic oxidation products in the atmosphere is further reduced (i.e. through gas or particle-phase chemistry; Jimenez et al., 2009; Donahue et al., 2011; Ehn et al., 2014), they may condense kinetically according to the Fuchs-corrected surface area of existing particles and a larger proportion of the condensable mass will be added to the nucleation mode (Riipinen et al., 2011; Yu, 2011; Zhang et al., 2012).
Neither approach fully describes the behaviour of SOA; the kinetic approach neglects the re-evaporation of semi-volatile organics whilst the thermodynamic approach is unable to account for the observed growth of particles beyond the nucleation mode (Pierce et al., 2011, 2012; Riipinen et al., 2011; Yu, 2011). The results of laboratory experiments indicate the presence of both semi-volatile organic aerosol components that evaporate upon heating or dilution (e.g. Robinson et al., 2007) as well as highly oxidised compounds with extremely low volatilities (Ehn et al., 2014) in atmospherically relevant SOA particles.
The treatment of SOA in global aerosol models is summarised by Tsigaridis et al. (2014). Global aerosol models typically use either the thermodynamic (partitioning proportional to existing organic mass; e.g. Chung and Seinfeld, 2002; Heald et al., 2008; Pye and Seinfeld, 2010; O'Donnell et al., 2011) or the kinetic (condensation proportional to particle surface area; e.g. Spracklen et al., 2006; Makkonen et al., 2009, 2012; Mann et al., 2010; D'Andrea et al., 2015) assumptions described above. Riipinen et al. (2011) and D'Andrea et al. (2013) both found that the simulated global annual mean concentration of CCN-sized particles increased by approximately 10 % when the kinetic (rather than mass-based) assumption was used for SOA, with regional increases of over 50 %. Yu (2011) found that allowing successive stages of oxidation to occur, and the generation of non-volatile products which were distributed according to the kinetic approach, increased simulated surface level CCN concentrations by 5–50 % over a version of the same model in which only an equilibrium approach was taken.
Previous modelling studies have quantified the cloud albedo effect, or first
AIE, of biogenic SOA, estimating global annual mean values that span from
positive (
Conversely, we hypothesise that the direct radiative effect (DRE) due to
biogenic SOA is less sensitive to the way in which secondary organic material
is partitioned across the size distribution. Previous studies agree on a
negative global annual mean DRE due to biogenic SOA:
Here we use a global aerosol microphysics model and offline radiative transfer model to explore how the assumed behaviour of secondary organic material affects simulated changes in CCN-sized particles and, for the first time, CDNC and the radiative effects of biogenic SOA.
We use GLOMAP-mode (Mann et al.,
2010), which is an extension to the TOMCAT model (Chipperfield, 2006). Here we
use the model with a horizontal resolution of
2.8
Global annual mean fluxes of secondary organic material to the
aerosol size distribution in GLOMAP.
We prescribe 6-hourly mean offline oxidant (OH, O
All model experiments include a representation of binary homogeneous
nucleation, which simulates the formation of particles from
H
The rate of change of gas-phase molecular concentration of organics
(
GLOMAP-mode uses a two-moment modal approach, within which mode number and mass concentrations vary but the mode width is held fixed. Following condensation, the particle number concentration for a mode is held constant and the new geometric mean diameter for the mode is calculated according to the updated volume per particle (Mann et al., 2010). Since the mode width is held fixed, this approach will not capture potential changes to the distribution of material within a mode following condensation.
Because our aim is to quantify the impact of changes in the size of particles to which the SOA condenses, we otherwise treat SOA identically between the two different approaches and do not allow SOA to re-partition into the gas phase. To isolate the impact of the condensation of secondary organic material the SOA yield, the oxidant chemistry and the SOA condensation rate was identical in all simulations.
The changes to CDNC due to the inclusion of biogenic SOA (i.e.
KinSOA
The first AIE due to biogenic SOA is then determined using the offline
radiative transfer model of Edwards and Slingo (1996) with nine bands in the
longwave (LW) and six bands in the shortwave (SW); a monthly mean climatology based on
ECMWF reanalysis data and cloud fields for the year 2000 from the ISCCP-D2
archive (Rossow and Schiffer, 1999) is used. To calculate the first AIE, a
uniform control cloud droplet effective radius (
To determine the DRE, following the methodology described in Rap et al. (2013), the radiative transfer model is used to calculate the difference in net top-of-atmosphere all-sky radiative flux between experiments including SOA and the equivalent experiments without SOA. Aerosol optical properties are calculated from the monthly mean aerosol distribution. The refractive index for each mode is calculated as the volume-weighted mean refractive index of the components present (values for which are given at 550 nm in Table A1 of Bellouin et al., 2011), including water for the soluble modes. The optical properties (scattering and absorption coefficients together with the dimensionless asymmetry parameter) are then obtained from look-up tables of all realistic combinations of refractive index and Mie parameter, as described in Bellouin et al. (2013).
Simulated and observed mean aerosol number size distributions at two
forested sites. (Left) Hyytiälä, Finland, during June–July–August
(multi-annual; 1996–2006). (Right) Manaus, Brazil, during the wet season
(multi-annual; January–June, 2008–2009). The grey shaded regions indicate
Table 1 summarises how the treatment of SOA condensation in GLOMAP impacts
the global flux of secondary organic material to the aerosol size
distribution. The total global flux of secondary organic material to the
particle phase is the same in each approach
(
Figure 1 compares simulated and observed aerosol number size distributions at
two forest locations: Hyytiälä, Finland (24.3
When the mass-based approach (dashed lines in Fig. 1) is applied, the model simulates high concentrations of nucleation mode particles (smaller than 10 nm dry diameter). Under this approach no organic material condenses on these small particles and their growth is slow since it is limited by the availability of sulfuric acid (Riipinen et al., 2011). This slow growth rate means few particles can grow to larger sizes and the number of particles simulated between 30 and 100 nm diameter is far less than observed at both sites.
When the kinetic approach is applied, a small amount of SOA condenses onto particles in the nucleation mode (0.26 % of the total global flux for ACT and 0.57 % for ORG; Table 1), increasing the growth rate of these small particles. At Hyytiälä, the kinetic approach results in a greater number of particles in the 30–100 nm size range compared to the mass-based approach, improving the match against observations over this size range (KinSOA; full lines in Fig. 1). At Manaus, the number of particles in the 30–100 nm range is underestimated with either approach. Under the kinetic approach, the number of particles between 100 and 200 nm is greater than observed, when the organically mediated new particle formation mechanism is used (KinSOA_ORG), suggesting that the growth rate of particles in the Aitken mode may be overestimated under these conditions.
Global annual mean change to CDNC, calculated using a globally
uniform updraught velocity of 0.2 m s
At both sites, we simulate a similar size distribution when we assume either completely kinetic condensation or a combination of the kinetic and mass-based approaches (full or dotted red lines in Fig. 1). In the SplitSOA experiments, some secondary organic material is able to condense onto newly formed particles (0.19 % of the total flux for ACT and 0.44 % for ORG; Table 1), resulting in a size distribution similar to that generated by the kinetic approach (SplitSOA; dotted lines in Fig. 1). This implies that it is not necessary to assume that all of the organic material generated has an extremely low volatility in order to account for the growth of particles to CCN relevant sizes, but that some fraction of the SOA does need to be treated in this way, consistent with the findings of Kulmala et al. (2013). The same relative response to the kinetic and mass-based approaches was documented by D'Andrea et al. (2013) and is consistent with Yu (2011), who combined the kinetic condensation of low-volatility organics with the equilibrium partitioning of semi-volatile organics.
Annual mean percentage change to CDNC (using a uniform updraught
velocity of 0.2 m s
Table 2 reports the simulated changes to global CDNC due to biogenic SOA
under the kinetic, mass-based and split approaches for both nucleation
mechanisms. As we have reported previously, the impact of biogenic SOA on
CDNC strongly depends on the nucleation mechanism used in the model, with the
largest impact on CDNC simulated with an organically mediated new particle
formation mechanism (Scott et al., 2014). We find that for both nucleation
mechanisms, biogenic SOA leads to a larger global annual mean increase in
CDNC under the kinetic condensation approach compared to the mass-based
approach. With the ACT nucleation mechanism, GLOMAP simulates a 3.4 %
increase in global annual mean CDNC with kinetic condensation but only a
1.4 % increase when the mass-based approach is used. Similarly, with the
ORG nucleation mechanism, GLOMAP simulates a 21.1 % increase in global
annual mean CDNC with kinetic condensation and a 16.8 % increase when the
mass-based approach is used; the same relative response is observed in our
sensitivity study (ORG_x10), in which the organically mediated nucleation
rate is a factor of 10 higher. The SplitSOA approach increases the global
annual mean CDNC by almost as much as the kinetic approach (
Simulated changes to CDNC (Fig. 2) are not restricted to the areas over which biogenic SOA is generated or proportional to the amount of SOA being generated, as discussed in Scott et al. (2014). Under the kinetic approach, GLOMAP simulates large fractional increases to CDNC over boreal regions and southern hemispheric oceans (greater than 20 % for ACT and over 100 % for ORG; Fig. 2). Using the mass-based approach, the growth of particles to CCN-relevant size is suppressed and fewer regions experience such large increases in CDNC in response to the inclusion of biogenic SOA.
The inclusion of biogenic SOA can induce decreases in simulated CDNC due to an enhanced rate of ageing/scavenging of initially hydrophobic particles or the suppression of nucleation in the free troposphere (and subsequent entrainment back into the boundary layer) due to an increased condensation sink at the surface (Scott et al., 2014). When the ACT mechanism for new particle formation is used with the kinetic approach, these effects combine to give small decreases to CDNC over tropical ocean regions (Fig. 2a); when the mass-based approach is used, much larger ocean regions experience a decrease in CDNC (Fig. 2b). In the ORG simulations, the contribution of organic oxidation products to new particle formation is sufficient to prevent any regional decreases in CDNC under either approach (Fig. 2c and d).
To examine the sensitivity of this response to updraught velocity, CDNC were
calculated at five globally uniform updraught velocities, from 0.1 to
0.5 m s
Annual mean first AIE (W m
We explored the impact of the condensation approach on the first AIE due to
biogenic SOA and report the global annual mean values in Table 2; Fig. 3
shows the spatial patterns in annual mean first AIE for each nucleation
mechanism and partitioning approach. The global annual mean first AIE is
calculated for an updraft velocity of 0.2 m s
Increases in CDNC lead to a negative first AIE (i.e. a cooling effect), and
decreases in CDNC lead to a positive first AIE (i.e. a warming effect). When
the ACT new particle formation mechanism is used, the kinetic approach gives
a negative global annual mean first AIE of
When the ORG mechanism is used with the kinetic approach, the inclusion of
biogenic SOA generates a global annual mean first AIE of
For both nucleation mechanisms, the SplitSOA approach gives a negative first
AIE (
Global annual mean change to cloud droplet number concentration (CDNC), calculated using five globally uniform updraught velocities, in the model level which corresponds to low-level cloud base (mean pressure of approximately 900 hPa), and first aerosol indirect effect (AIE), reported to two decimal places, resulting from the inclusion of biogenic SOA in GLOMAP using the kinetic and mass-based approaches. The mean values are calculated assuming that all updraught velocities are equally likely.
The direct radiative effect is less sensitive to the distribution of
secondary organic material across the aerosol size distribution (Table 2).
When the ACT new particle formation mechanism is used, the kinetic approach
gives a global annual mean DRE of
This lack of sensitivity can be explained in part by the fact that, under the
mass-based approach, the shift in SOA mass from smaller (
Using a global aerosol microphysics model we have shown that the manner in which secondary organic material is added to the aerosol size distribution is important in determining its impact on the number and size of climatically relevant particles in the atmosphere. We tested two assumptions about the way secondary organic material is added to existing particles – a kinetic approach where material condenses according to the surface area of existing particles and a mass-based approach where the material condenses according to the organic mass present in existing particles. We calculate the impact of these assumptions on the direct radiative effect and first aerosol indirect effect due to biogenic secondary organic aerosol.
The kinetic approach to partitioning, which enables organic oxidation
products to condense upon the smallest particles, facilitating their growth
to larger sizes, increases global annual mean CDNC when either an activation
or organically mediated new particle formation mechanism is applied (by
3.4 % for ACT and 21.1 % for ORG). These global annual mean increases
in CDNC result in a negative global annual mean first AIE
(
Applying a mass-based approach to partitioning suppresses the growth of the
smallest particles, resulting in a smaller global annual mean increase in
CDNC (
Neither approach adequately describes the complex behaviour of secondary organic material in the atmosphere. Ultimately, a combination of thermodynamic and kinetic approaches will be required to accurately represent the behaviour of SOA in global models. Improving on existing first attempts, by e.g. Riipinen et al. (2011), Yu (2011) and the SplitSOA approach described here, will require a more detailed understanding of the pathways by which organic compounds of differing volatilities are generated and their relative contributions to the growth of particles of different sizes.
We acknowledge support from NERC (NE/H524673/1, NE/J004723/1, NE/G015015/1, NE/K015966/1), EPSRC (EP/I014721/1), ERC (227463-ATMNUCLE). Aerosol measurements at Hyytiälä were supported by the Academy of Finland Centre of Excellence (1118615 and 1127372), and the Cryosphere-Atmosphere Interactions in a Changing Arctic Climate (CRAICC) programme. Aerosol measurements in the Amazon were supported by Fundacao de Amparo a Pesquisa do Estado de Sao Paulo (FAPESP – AEROCLIMA 08/58100-2), Conselho Nacional de Desenvolvimento Científico (CNPq) and European Integrated FP6 Project on Aerosol Cloud Climate and Air Quality Interactions (EUCAARI – 34684), under the scope of LBA experiment; we thank INPA (Instituto Nacional de Pesquisas da Amazonia) for the coordination work of the LBA Experiment. Edited by: K. Tsigaridis