The gas-particle partitioning and chemical aging of semi-volatile organic aerosol are presented in a newly developed box model scheme, where its effect on the growth, composition, and mixing state of particles is examined. The volatility-basis set (VBS) framework is implemented into the aerosol microphysical scheme MATRIX (Multiconfiguration Aerosol TRacker of mIXing state), which resolves mass and number aerosol concentrations and in multiple mixing-state classes. The new scheme, MATRIX-VBS, has the potential to significantly advance the representation of organic aerosols in Earth system models by improving upon the conventional representation as non-volatile particulate organic matter, often also with an assumed fixed size distribution. We present results from idealized cases representing Beijing, Mexico City, a Finnish forest, and a southeastern US forest, and investigate the evolution of mass concentrations and volatility distributions for organic species across the gas and particle phases, as well as assessing their mixing state among aerosol populations. Emitted semi-volatile primary organic aerosols evaporate almost completely in the intermediate-volatility range, while they remain in the particle phase in the low-volatility range. Their volatility distribution at any point in time depends on the applied emission factors, oxidation by OH radicals, and temperature. We also compare against parallel simulations with the original scheme, which represented only the particulate and non-volatile component of the organic aerosol, examining how differently the condensed-phase organic matter is distributed across the mixing states in the model. The results demonstrate the importance of representing organic aerosol as a semi-volatile aerosol, and explicitly calculating the partitioning of organic species between the gas and particulate phases.
Atmospheric aerosols play a key role in the Earth system with great impacts on global air quality, public health, and climate (Boucher et al., 2013; Myhre et al., 2013; Seinfeld and Pandis, 2016). One contribution to the large uncertainty in aerosol radiative forcing is organic aerosol (OA), which is ubiquitous in the atmosphere and contributes to a large portion of submicron particulate mass in various regions around the world (Zhang et al., 2007; Jimenez et al., 2009). Advancements in measurement techniques greatly improved our understanding of the evolution of OA and its lifetime in the atmosphere at the process level (Jimenez et al., 2009). However, OA processes in models still remain poorly constrained. Measurements imply that OA concentrations are potentially underestimated in current models (Tsigaridis et al., 2014). Such a discrepancy hints at large uncertainties in the prediction of aerosol–radiation interactions, their hygroscopicity, aerosol–cloud interactions, and their overall impact on climate (Petters and Kreidenweis, 2007).
Missing sources of secondary organic aerosol (SOA) in models have been suggested to be the main cause of the underestimated OA formation (Heald et al., 2005; Volkamer et al., 2006; Hodzic et al., 2010; Spracklen et al., 2011). More recently, studies have sought to investigate the underestimation of organic aerosol mass within more advanced model frameworks, which are capable of resolving semi-volatile primary organic aerosol (POA) and including SOA from a wider set of precursors including intermediate-volatility organic compounds (IVOCs). The volatility-basis set (VBS) was developed (Donahue et al., 2006) to provide a relatively simple framework whereby models can represent the overall behavior of the myriad of compounds that constitute organic aerosol and their precursors. The approach involves considering OA as being composed of a number of representative species, each with a particular volatility, spanning a spectrum in vapor pressures from highly volatile (which essentially remains in the gas phase) to very low vapor pressure species which partition readily into the particle phase. VBS then captures the chemical aging of the organic species in the gas phase, with the hydroxyl radical oxidizing them and producing the adjacent lower-volatility class as a product. This method has been used extensively in regional studies (Robinson et al., 2007; Shrivastava et al., 2008; Murphy and Pandis, 2009; Tsimpidi et al., 2010, 2011; Hodzic et al., 2010; Fountoukis et al., 2011, 2014; Bergström et al., 2012; Athanasopoulou et al., 2013; Zhang et al., 2013) but less so in global models (Pye and Seinfeld, 2010; Jathar et al., 2011; Jo et al., 2013; Tsimpidi et al., 2014; Hodzic et al., 2016). Other studies have used the 2-D VBS (Donahue et al., 2011; Murphy et al., 2011), an approach that, in addition to the volatility space, also resolves that of chemical composition, by tracking the amount of oxygenation in the representative organic compounds. However, the 2-D VBS is not implemented in global models, due to its large number of tracers and the large number of free parameters that are involved in the parameterization.
Naming convention and parameters used in the VBS implementation described here.
The inclusion of semi-volatile organics is important for accounting for the total mass of organics in the particulate phase, since an increase in particulate organic matter may not be the result of chemically produced low-volatility species, but simply be reflecting a temperature-driven increase in the partitioning of semi-volatile organic aerosol into the particle phase. It has been established that the highly oxidized, very low volatility organics play a key role in particle formation (Metzger et al., 2010; Paasonen et al., 2013; Riccobono et al., 2014; Kirkby et al., 2016) and particle growth (Tröstl et al., 2016), while the range of volatilities contributing to aerosol growth increases with aerosol size (Pierce et al., 2011; Yu, 2011). Semi-volatile organics also affect aerosol size and mixing state, as well as their impact on climate, due to changes in cloud condensation nuclei (CCN) formation rates (Petters et al., 2006, Riipenen et al., 2011; Scott et al., 2015), hygroscopicity (Petters and Kreidenweis, 2007), and optical properties (Myhre et al., 2013). Since OA emissions are on the rise from developing countries (Lamarque et al., 2010) and no Earth system model considers anthropogenic OA as semi-volatile as measurements suggest, it is important to include and constrain semi-volatile organics to ultimately reduce uncertainties in aerosol radiative forcing and make climate-model-simulated aerosol changes more realistic.
The objective of this study is to further develop an aerosol microphysics model by including a more advanced representation of organic aerosol, including semi-volatile primary OA and an evolving OA volatility during chemical aging in the gas phase, in its calculations. This objective was achieved by implementing the VBS framework in the aerosol microphysical scheme MATRIX (Multiconfiguration Aerosol TRacker of mIXing state; Bauer et al., 2008), which represents major aerosol processes such as nucleation, condensation (excluding organics in its original version), and coagulation, and explicitly tracks the mixing state of different aerosol populations. As many traditional chemistry–climate models do (Tsigaridis et al., 2014), MATRIX treats POA and SOA as non-volatile (Bauer et al., 2008). By coupling MATRIX with VBS, POA are treated as condensable semi-volatile organic compounds. These can partition among different aerosol populations based on their volatility and aerosol population size distribution, capturing particle growth via condensation of low-volatility organic vapors and thus providing a more physically based calculation of aerosol microphysics.
A box model is used for this study. The gas-phase chemical mechanism CBM-IV (Carbon Bond Mechanism IV; Gery et al., 1989), as used in the NASA GISS ModelE (Shindell et al., 2001, 2003), is coupled to the MATRIX aerosol microphysics scheme, utilizing the Kinetic Pre-Processor (KPP; Sandu and Sander, 2006) to solve the differential equations of the gas-phase chemistry scheme. A time step of 30 min is used, for consistency with the global model.
MATRIX (Bauer et al., 2008) is an aerosol microphysical model based on the quadrature method of moments (McGraw, 1997) in the NASA GISS ModelE Earth system model,
which can be used either as a module within the global model or as a
stand-alone box model. Here, the stand-alone box model is used for
development. The design of the code is such that the box-model code can be
used as-is in the global model, without any changes, allowing for seamless
transition and maximum portability. MATRIX is designed to resolve the
aerosol temporal evolution and represent the mixing states of a
user-selected set of aerosol populations, which are modes of different
composition as listed in Table 2, tracking two moments each, number and
mass, while keeping the width of the distribution fixed. It describes new
particle formation, particle growth through condensation with explicit
treatment of sulfuric acid condensation, and lumped treatment of the
NH
Black carbon is uniquely treated in MATRIX, in order to separate the coated (via condensation) from the mixed (via coagulation) populations. It is emitted in BC1, which can grow (blue arrow in Fig. 1) with inorganic and organic coating, and as its coating volume fraction reaches 5 %, it would be moved in the BC2 population (see Table 2 for a description), shown as the orange circle.
Schematic showing coagulation pathways among organics-containing aerosol populations as colored circles, with nine organic VBS species condensed as grey outer circles. In yellow are the emitted donor aerosol populations, and in green are the mixed recipient populations. OCC has a semi-transparent yellow core because it is actually emitted as the VBS species that can serve as condensation medium for gaseous VBS species, represented by the grey outer circles. In orange is population BC2, which contains a > 5 % coating of sulfate and organics, which is formed rapidly from the growth of population BC1, which has < 5 % sulfate and/or organics coating.
The volatility-basis set approach is introduced to the original model; it is
an organic aerosol volatility parameterization that separates semi-volatile
organic compounds into logarithmically spaced bins of effective saturation
concentrations, which are used for gas-particle partitioning and
photochemical aging (Donahue et al., 2006). The scheme groups organic
compounds into nine surrogate VBS species according to their effective
saturation concentrations (
Aerosol population chemical composition in MATRIX.
Note: The sigma values for all populations are 1.80, except for AKK, which has a sigma of 1.60, and for SSC and MXX, which both have a sigma of 2.00.
The emission rates for the VBS species were derived from the POA emission rate in the global model for the corresponding grid box and month, which were distributed in the volatility space by using mass-based emission factors from Shrivastava et al. (2008) (Table 1). Adding up the 9 factors from each bin listed in Table 1, we obtain a total factor of 2.5, which means the new scheme's organics emission is 2.5 times that of the organics emissions in the original scheme. The additional multiplication factor of 1.5 is applied to the emission to account for missing sources of volatile organics in the IVOC volatility regime in the inventories (Shrivastava et al., 2008).
Conditions of each location used in the simulations, taken from the GISS ModelE for January and July 2006.
* The new scheme's organics emissions are 2.5 times higher than the organics emissions in the original scheme
In the original version of the MATRIX model, organics only contribute to particle growth and mix with other aerosol species via coagulation. Primary organic aerosols are emitted only as non-volatile particulate organic matter, and do not exist in the gas phase or interact with other aerosol populations. Implementing the VBS scheme adds these missing processes. Before this development, there were eight alternative configurations of MATRIX available to the user, each representing a distinct set of aerosol populations whose number, composition, and interactions by coagulation vary. A 9th configuration with 15 selected aerosol populations is created for this study (Table 2), in which 8 of the 16 populations, ACC, OCC, BC1, BC2, OCS, BOC, BCS, and MXX, could contain organics as semi-volatile VBS species. We only included semi-volatile organics in eight populations, so that we can examine the BC–OA–sulfate–nitrate system first, before adding them into the nucleation population AKK and the dust and sea salt populations (DD1, DS1, DD2, DS2, SSA, SSC). Through coagulation, the 15 donor populations grow or mix and are placed into recipient populations, based on the donor population composition, as described in Bauer et al. (2008). In a future stage, organics will also be implemented in the AKK mode to present nanoparticle growth, and we will include an additional nucleation scheme that considers the dependence of new particle formation that involve organics (Kirkby et al., 2016; Tröstl et al., 2016).
Temporal evolution of the mass concentration of semi-volatile organics in the gas phase (left column), aerosol phase (across all populations; middle column) and total (right column) using the new scheme (refer to Table 1 for legend) for January. The total of non-volatile organics from the original scheme (OCAR) is shown in black dash-dotted lines in the aerosol phase column (middle). OCAR from the old scheme is exactly 2.5 times smaller from the total organic species in the new scheme.
Same as Fig. 2, for July.
Previously, each aerosol population carried up to five tracers – sulfate, black carbon, non-volatile organics, dust, and sea salt. Now each of the eight organics-containing populations carry nine additional semi-volatile VBS species listed in Table 1. Together with the 5 original tracers, we now have up to 14 available tracers per population, depending on whether they carry organic aerosols or not, with the original organics tracer (OCAR) representing the non-volatile biogenic OA, as it did in the original mechanism. This newly coupled model MATRIX–VBS treats POA as semi-volatile gas-phase species, which then partition into and out of the particulate phase. The amount of gas-phase species partitioned onto each aerosol population is based on the surface area of that population, in addition to the mass of that population and the volatility of species, and equilibrium partitioning is assumed.
The semi-volatile nature of biogenic SOA is not represented in the VBS framework in this work. Instead, biogenic SOA are treated as non-volatile, as in the original MATRIX version, and are produced with a 10 % constant yield from terpenes emissions without any requirement for oxidation before the OA is formed (Lathière et al., 2005; Tsigaridis et al., 2014). The inclusion of semi-volatile biogenic SOA will be parameterized in the same way as in the VBS framework presented here in the future.
To test the newly developed model's behavior, we simulated idealized cases representative of four different locations and environments: one very polluted city (Beijing), another cleaner yet still very polluted city at high altitude and closer to the tropics (Mexico City), a very clean Finnish forest (Hyytiälä), and an anthropogenically affected forest in the southeast USA (Centreville, Alabama). The experiments are performed for a winter month (January) and a summer month (July) for 10 days, and initial conditions and emission rates for each location were extracted from a GISS ModelE simulation (similar setup as described in Mezuman et al., 2016) for the year 2006, listed in Table 3. All parameters and emissions are held constant throughout the simulations. Here we do not include deposition and dilution, for simpler mass-balance calculations. Semi-volatile POA, sulfate in the accumulation mode, and black carbon are emitted continuously in the OCC, ACC, and BC1 populations, respectively, shown in Fig. 1 as yellow circles. Condensation of VBS species on BC1 can increase the non-absorbing shell of that population, leading to the formation of BC2, as described above.
The four organics-containing populations described above can coagulate (black arrows in Fig. 1) with themselves and each other and form three additional organics-containing mixed populations, BOC, OCS, and BCS, shown as green circles. This schematic includes seven of the eight organics-containing populations in the model.
The temporal evolution of the total organics mass concentrations from the
new scheme and the old scheme are presented in Fig. 2 for January and
Fig. 3 for July in the four locations under study. They show large changes
in organics concentrations between the old scheme (black line on the right
column) and the new one (colors). The organics in the new scheme are
represented and distributed by organics tracers of different volatility,
whose saturation concentration
As mentioned in the model description, the emission rates for organics in each of the volatility bins in the new scheme were derived from the Shrivastava et al. (2008) mass-based emission factors. Consequently, since there is no deposition and dilution in the simulations, the new scheme's organics total mass concentrations (shown in color in the right columns of Figs. 2 and 3) always adds up to 2.5 times that of the old scheme (shown as dash-dotted lines) throughout the simulations in both January and July.
In January, the total mass concentration for organics in Beijing,
Centreville, Hyytiälä, and Mexico City at the end of 10 days are
approximately 115, 16, 13,
and 65
In all four locations, almost all species in the intermediate-volatility range are in the gas phase, those in the semi-volatile range partition between the gas and aerosol phases, and those in the low-volatility range are in the aerosol phase in January. This is especially true for Beijing and Hyytiälä, where the volatility distributions are very similar (in relative terms) and where the total concentration of gas-phase species is higher than the sum of all aerosol-phase species. In Centreville, the total amount of gas-phase species is approximately the same as that of the aerosol-phase species, whereas in Mexico City there are more species in the aerosol phase than in the gas phase. In Centreville and Mexico City, the species show a diurnal variability, which will be explained later.
Aging can help explain the similar volatility distributions in Beijing and
Hyytiälä. The
Looking at the total of the organics (right column of Fig. 2), it is not
surprising that the very polluted Beijing has the highest concentration of
total organics while the cleanest location, Hyytiälä, has the
lowest; what is interesting, however, is that organics at these locations
share similar volatility distributions. By the end of the 10-day simulations
in the new scheme, the volatility distributions in Beijing and
Hyytiälä are very similar to the emission factor distribution among
the volatility, with factor differences of less than 0.1. This behavior is,
again, a result of the low
On the other hand, the volatility distributions in Centreville and Mexico
City are very different from the applied emission factor distribution,
except for the two bins in the low-volatility range, M2 and M1. Due to the high
concentrations of
The total mass concentration of organics in Beijing and Mexico City at the
end of 10 days in July are approximately 130 and 67
Temporal evolution of organic aerosol mass concentration in each organics-containing population from the new scheme (first column for January, third column for July), and the old scheme (second column for January, fourth column for July).
The temporal evolution of total organic aerosol mass concentration per population is shown in Fig. 4 (absolute amounts) and Fig. 5 (relative amounts). The first and third columns are results from the new scheme with condensing and coagulating organics for January and July, respectively, while the second and fourth columns are results from the old scheme with only coagulating organics from January and July, respectively. The organic aerosol mass concentrations in Fig. 4 correspond to the aerosol-phase concentrations in Figs. 2 and 3 (middle column), except they are now separated by population, whereas in the earlier two figures they were separated by mass tracers representing volatility. At a first glance, the population with the highest organic mass is BOC for January and July in both schemes. BOC is the population that contains OC, BC, and sulfate, and is the end result of coagulation of all populations in our idealized cases. However, in the old scheme, populations OCC and OCS also have significant amounts of organics. This is because in the new scheme the emitted populations are ACC, BC1, and OCC, and organics that are emitted in the OCC population can condense on and/or coagulate with other populations, including being lost by evaporation and then repartitioning to other populations. Thus there is an additional loss mechanism of organics from those populations in the new scheme. In addition, there is competition between the ACC and BC1 populations in both schemes, and in the new scheme, aerosol-phase organics in the OCC population could either coagulate with the ACC population to form OCS, or they could coagulate with the BC1 population to form BOC. This competition determines how much OCS and BOC are formed, and it affects how much gas-phase organics from the OCC population could condense on the two populations and the distribution of organics among the populations. Since partitioning adds a loss mechanism to OCC, part of the evaporated mass will go to BOC, making it larger, and a more efficient scavenger of other particles. As a result, most organics coagulate with and condense on the BOC population and/or the OCS population, and together with the emitted OCC population, hold the most organics and dominate the mass fractions.
Temporal evolution of organic aerosol mass concentration fraction in each organics-containing population from the new scheme (first column for January, third column for July), and the old scheme (second column for January, fourth column for July).
There is some similarity between the January and July results between the
new and the old schemes (Fig. 5). This similarity means that the
distribution of organics among aerosol populations is not significantly
affected by season. This is consistent with a study by Bauer et al. (2013),
where they found that the mixing-state distribution is rather a
characteristic of a region and not so much of a season, although the total
(absolute) amounts by season may vary. By the end of the simulations, most
locations have more organics present in the BOC population, except those in
Centreville. The reason for this is sulfate; from the sulfate and black
carbon emissions listed in Table 3, we can calculate the sulfate-to-black-carbon ratio in Centreville to be
Another important factor on the evolution of aerosols is their size distribution. Shown in Figs. 6 and 7 are the January size distributions from Mexico City and Centreville, respectively. The first row shows number concentration, the second row surface area, and the third row volume. The first two columns are results from the new and old schemes after 24 h of simulation, and the right two columns are after 120 h. The total number concentration, surface area, and volume from the eight populations are shown as dotted lines. Note also that these plots show the total aerosol size distribution per population, which includes the contribution of species other than organics.
Organics-containing aerosol populations (except MXX) and AKK (Aitken mode sulfate) size distributions for Mexico City in January. Top row: number concentration. Middle row: surface area. Bottom row: volume. Total of all populations in dotted black lines.
Same as Fig. 6 for Centreville.
The size distributions in July are very similar to January in all locations, and therefore only January is shown here. Beijing, Hyytiälä, and Mexico City exhibit somewhat similar size distributions (with different absolute amounts), just as their mass fractions do. The size distribution is dominated by OCC, OCS, and BOC in the first 3 to 4 days, but later only by BOC. On the other hand, Centreville, similar to its mixing state, is different in its size distribution of different aerosol populations from the other three locations. Therefore, only the size distributions of Mexico City and Centreville are shown here.
In the new scheme for Mexico City after 24 h of simulation, the number
concentration has two modes. OCC has even smaller size than Aitken mode
sulfate AKK does, as a result of the evaporation of organics, but its number
concentration is higher. OCS and BOC have started to form from coagulation
of OCC with ACC and BC1, and their diameter, number concentration, surface
area, and volume are very similar, almost overlapping, with BOC slightly
smaller in diameter. After 120 h of simulation, OCC's number
concentration has decreased significantly, from
In the old scheme, after 24 h OCC has higher number concentration
(peaking at
The Centreville size distributions tell a different story. In the early
stages of the new scheme, OCS has greater number concentration and size
than BOC does; OCS's peak number concentration is
Organic aerosol volatility calculations were implemented into a new aerosol
microphysics scheme, MATRIX-VBS. Results from idealized cases in Beijing,
Centreville, Hyytiälä, and Mexico City during summer and winter using
the new scheme were compared against the original scheme and showed how the
inclusion of semi-volatility of organics and their reactivity affected
aerosol mass concentration, as well as their mixing state and size
distribution. Emission factors,
Going forward, the new scheme will be simplified, and we will reduce the number of tracers needed in order to simplify the model and save computational resources, without losing the essential information needed for volatility. The simplified version of the box model will then be implemented in the NASA GISS ModelE Earth system model. While this study is purely theoretical, we will evaluate MATRIX-VBS after its implementation into GISS ModelE. We will gain even better understanding of how semi-volatile organics are altering aerosol mixing state, how meteorological conditions and pollution levels influence organics' volatility distribution (as well as their mixing state in the real world), and what implications these processes have on the climate system.
This model development is part of GISS ModelE Earth system model, which is publicly available. In addition, the box model used here is available on request.
The authors declare that they have no conflict of interest.
We thank the NASA Modeling, Analysis, and Prediction program, which supports the GISS ModelE development. Edited by: G. Mann Reviewed by: three anonymous referees