Sensitivity of a global model to the uptake of N 2 O 5 by tropospheric aerosol

The uptake of N2O5 on aerosol impacts atmospheric concentrations of NO x and so O3, OH, and hence CH4. Laboratory studies show significant variation in the rate of uptake, with a general decline in the value of γN2O5 over the last decade as increasingly relevant tropospheric proxies have been studied. In order to understand the implication of this decline for tropospheric composition, a global model of tropospheric chemistry and transport (GEOS-Chem) is run with differing values of γN2O5 (0.0, 10−6, 10−4, 10−3, 5×10−3, 10−2, 2×10−2, 0.1, 0.2, 0.5, and 1.0). We identify three regimes in the model response. At low values ofγN2O5, the model shows reduced sensitivity to the value ofγN2O5 as heterogeneous uptake of N 2O5 does not provide a significant pathway to perturb NO x burdens. At high values ofγN2O5 the model again shows reduced sensitivity to the value ofγN2O5, as NOx loss through heterogeneous removal of N2O5 is limited by the rate of production of NO3, rather than the rate of heterogeneous uptake. At intermediate values of γN2O5 the model shows significant sensitivity to the value ofγN2O5. We find regional differences in the model’s response to changing γN2O5. Regions with high aerosol surface area and low temperatures show NO 3 production becoming rate limiting at lower γN2O5 values than regions with lower aerosol surface area and higher temperatures. The northern extra-tropics show significant sensitivity to the value ofγN2O5 at values consistent with current literature (0.001–0.02), thus an accurate description of γN2O5 is required for adequate simulation of O 3 burdens and longrange transport of pollutants in this region. Our model simulations also provide insight into the sensitivity of coupled chemistry-aerosol simulations to the choice of γN2O5. We find little change in the global sensitivity of NOx, O3 and OH toγN2O5 in the range 0.05 to 1.0, but a Correspondence to: H. L. Macintyre (h.macintyre@see.leeds.ac.uk) significant drop in sensitivity below this range. Thus simulations of the coupled impact of both chemistry and aerosol changes through time will be sensitive to the choice of γN2O5.


Introduction
Aerosols provide a significant source of uncertainty in our understanding of climate (Solomon et al., 2007).As well as affecting climate through direct and indirect radiative forcings (Lohmann and Feichter, 2005), they can impact photochemistry and thus the concentration of climate relevant gases such as methane (CH 4 ) and ozone (O 3 ).This impact is achieved both by changing the photolysis rate of species (Wild et al., 2000;Martin et al., 2003) and by providing a surface upon which heterogeneous reactions can occur (Dentener and Crutzen, 1993;Jacob, 2000).The dominant heterogeneous reaction in the troposphere is the reactive uptake of N 2 O 5 (Tie et al., 2001;Martin et al., 2003;Liao et al., 2003) producing nitric acid (HNO 3 ), thus removing oxides of nitrogen from the atmosphere.Laboratory studies show the rate of reactive uptake of N 2 O 5 varies with aerosol composition, temperature and relative humidity, with more recent studies tending to give lower values (Hanson and Ravishankara, 1991;Kane et al., 2001;Hallquist et al., 2003;Thornton et al., 2003;Badger et al., 2006;Brown et al., 2009;Bertram et al., 2009).In this paper the sensitivity of a global composition transport model to the uptake of N 2 O 5 is investigated to gauge the impact of this downward trend.
Tropospheric NO x (=NO+NO 2 ) sources are dominated by anthropogenic combustion processes, with significant natural sources including lightning and soils (Solomon et al., 2007).The conversion of NO x to HNO 3 is the most significant sink for NO x .During the day this is achieved by the reaction of NO 2 with OH (Reaction R1).At night NO 2 can react with O 3 to produce NO 3 , and then NO 3 can react with NO 2 to produce N 2 O 5 , which subsequently reacts on aerosol (Reactions R2-R4) to produce aerosol nitrate.This channel is only significant at night as NO 3 concentrations are low during the day due to its photolysis.
The significance of N 2 O 5 uptake as a NO x sink is shown both by modelling studies (Dentener and Crutzen, 1993;Tie et al., 2001;Evans and Jacob, 2005), and field evidence, (Platt et al., 1984;Munger et al., 1998;Brown et al., 2006).NO x concentrations impact the production of O 3 and thus the concentration of OH, one of the most important tropospheric oxidants (Logan et al., 1981).Thus understanding the sources and sinks of NO x is important for climate, as they impact the global O 3 and CH 4 burdens (the dominant sink for CH 4 is reaction with OH), and hence radiative forcing.
Model representation of the heterogeneous process is achieved by an uptake parameter, gamma (γ ), defined as the probability that a molecule impacting the surface of an aerosol undergoes irreversible reaction (Schwartz, 1986;Dentener and Crutzen, 1993).Initial laboratory studies to determine γ N 2 O 5 were performed on cold sulfuric acid aerosol as a proxy for stratospheric conditions (Mozurkewich and Calvert, 1988;Hanson and Ravishankara, 1991;Van Doren et al., 1991;Fried et al., 1994).Relatively high values of γ N 2 O 5 (∼0.1) were found.Given the lack of measurements of γ N 2 O 5 for tropospherically-relevant temperatures, humidities and aerosol compositions, this value was adopted for global tropospheric modelling.Dentener and Crutzen (1993) included reactive uptake in a global chemistry-transport model with a γ N 2 O 5 value of 0.1, and found a global reduction in NO x of 49%, with a corresponding drop in O 3 and OH of 9% each (examining only the winter Northern Hemisphere yields a reduction in NO x of 75%, with corresponding drops in O 3 and OH of 20% and 25%, respectively).Other modelling studies examining the effect of aerosol on tropospheric oxidants found surface reductions in NO x and O 3 of 80% and 10-30%, respectively, for winter northern latitudes (Tie et al., 2001(Tie et al., , 2003;;Liao et al., 2003).
As laboratory studies were performed for more tropospherically relevant aerosol, it was apparent that the value of γ N 2 O 5 from the early lab studies (∼0.1) was too high for the troposphere.Evans and Jacob (2005) used more appropriate laboratory studies to develop a new scheme for uptake more suited to the troposphere.This used data from studies on single component aerosol (neutralised sulfate, dust, organics and sea-salt).A global average γ N 2 O 5 of 0.02 was calculated, increasing the mass weighted global NO x , O 3 and OH burdens by 7%, 4% and 8%, respectively, compared to values simulated using the previous γ N 2 O 5 value of 0.1.
Recent studies suggest that γ N 2 O 5 values may be smaller still than those used in Evans and Jacob (2005).Brown et al. (2009) estimate the value on ambient aerosol to be in the range 5×10 −4 -6×10 −3 , and Bertram et al. (2009) in the range 3×10 −3 -9×10 −3 , both roughly a factor of ten lower than the mean value of 0.02 found by Evans and Jacob (2005).Both studies also characterise the aerosol and find up to 60% (Brown et al., 2009) and 75% (Bertram et al., 2009) consists of organic material.Throughout the troposphere, organic material makes up a significant and sometimes dominant fraction of aerosol mass (Zhang et al., 2007).Evidence from both field and lab studies suggests this organic material may form a coating (Pósfai et al., 1998;Tervahattu et al., 2002;Vaden et al., 2010).It has been shown that additional organic material can significantly reduce the uptake coefficient on sulfate aerosol, in some cases by several orders of magnitude (Folkers et al., 2003;Badger et al., 2006;Anttila et al., 2006;Brown et al., 2009), and that this coating need not be complete to suppress γ N 2 O 5 (Cosman and Bertram, 2008).Therefore the estimates for ambient aerosol from single-component studies made by Evans and Jacob (2005) may be too large.A study by Alexander et al. (2009) looked at modelled and observed 17 O to constrain nitrate formation pathways, and found that the model overestimated production through N 2 O 5 hydrolysis at winter northern latitudes, lending further support to the fact that γ N 2 O 5 may be lower than previously thought.
Recent studies have shown that ClNO 2 may also be produced by reaction of N 2 O 5 on chloride-containing aerosol (Osthoff et al., 2008;Roberts et al., 2009).This would act to reduce the nitrate source from the heterogeneous reaction, as well as reducing the NO x sink (ClNO 2 may photolyse producing NO 2 and a Cl radical).This could also be viewed as a lowering of the γ N 2 O 5 value for Reaction R4, as it similarly results in a smaller NO x sink and lower nitrate production.
Thus, heterogeneous uptake of N 2 O 5 is a significant driver of tropospheric composition.However there are a wide range of γ N 2 O 5 within the literature (10 −4 to >0.1), with recent assessments using real atmospheric aerosols suggesting even lower values than previously considered.Conceptually, below some value of γ N 2 O 5 , the heterogeneous uptake of N 2 O 5 becomes an insignificant process for the loss of NO x and therefore an precise definition of γ N 2 O 5 is not necessary.In this paper we investigate the sensitivity of a model of tropospheric chemistry and transport to a wide range of γ N 2 O 5 in order to investigate model response to different values of γ N 2 O 5 .

Model simulations
We use the GEOS-Chem global chemical-transport model version v8-01-04 (http://acmg.seas.harvard.edu/geos/)(Bey et al., 2001), driven by assimilated meteorological fields from the Goddard Earth Observing System (GEOS) of the NASA Global Modelling Assimilation Office (GMAO).For computational expedience, simulations are run at 4 • latitude by 5 • longitude, with 30 vertical layers.The model carries five externally-mixed aerosol types (sulfate, black carbon, organic carbon, dust, and sea-salt), with a relative humidity dependent size distribution based on Martin et al. (2003) and references therein.The model contains a detailed representation of NO x -O x -hydrocarbon chemistry, including reactions of NO 3 with the organic compounds isoprene, propene, dimethyl sulfide, formaldehyde, acetaldehyde, C>4 alkanes, C>2 aldehydes, C>3 ketones, glyoxal, methylglyoxal, methylvinylketone, methacrolein and ethane, and their subsequent degradation chemistry.In order to simplify the analysis, the value of γ N 2 O 5 is set to a single value across all aerosol types and ambient conditions for each simulation.γ N 2 O 5 values of 0.0, 10 −6 , 10 −4 , 10 −3 , 5×10 −3 , 10 −2 , 2×10 −2 , 0.1, 0.2, 0.5, and 1.0 are chosen to fill the range of possible values.First order loss rate coefficients are calculated using the equation of Schwartz (1986), and the heterogeneous reaction is assumed to proceed via Reaction R4.Each simulation is run for two years.The analysis is performed on the second year of output, and the first year discarded as spin-up.
The GEOS-Chem model has been extensively used and has previously been evaluated against observations for many locations (Bey et al., 2001;Martin et al., 2003;Evans and Jacob, 2005;Zhang et al., 2008;Nassar et al., 2009).At the resolution used here, correlations between tracers, (e.g.aerosol and NO x ) in plumes, may not be fully resolved, with subsequent impacts on chemistry.

Impact on concentrations
Figure 1 shows the impact that changing γ N 2 O 5 has on the monthly mean mass averaged burden of tropospheric NO x in various latitude bands, integrated up to the tropopause.The greatest impact is seen on concentrations in the northern extra-tropics during the winter months.This can be explained by the large NO x and aerosol sources in this region, coupled to the longer nights during the Northern Hemisphere winter.The low solar insolation in the northern extratropics winter leads to lower OH concentrations, and hence reduced rates of OH loss through the reaction of NO 2 with OH (Reaction R1).The long nights enhance NO 3 persistence and hence NO x loss through N 2 O 5 hydrolysis.Tropical regions do not exhibit this strong seasonality, but rather show a steady shift from the baseline concentration, due to the more consistent solar insolation year round.Considerable aerosol loadings in the tropical regions, mainly from biomass and biogenic sources, provide a potentially large sink for NO x .However, the persistence of NO 3 is inhibited by its photolysis (and also its reaction with abundant organic compounds) causing a much smaller overall impact on NO x concentrations compared to the winter northern extratropics.Despite the long nights in the southern extra-tropics, aerosol and NO x loadings are low here, hence little impact is seen.It is noted that for the region which displays the highest sensitivity (i.e.winter northern latitudes), there exist very few appropriate lab or field determinations of γ N 2 O 5 .It is likely that additional organic material in the aerosol (which lowers γ N 2 O 5 ) may not be as prevalent during this time (due to low productivity), therefore further determinations of the true value of γ N 2 O 5 for aerosol types that dominate this region would be beneficial.
Figure 2 shows the mean annual tropospheric burdens of NO x , O 3 and OH as a function of γ N 2 O 5 (shown on a log scale), for the northern extra-tropics, southern extra-tropics, tropical regions, and the whole globe.Examining the NO x burden in the northern extra-tropics, there are three obvious regimes.At high and low values of γ N 2 O 5 the model shows reduced sensitivity to the value of γ N 2 O 5 (perturbations in γ N 2 O 5 lead to very small changes in burdens) with a transitional regime where changes in γ N 2 O 5 lead to significant changes in the simulated burdens.The other regions and species show a similar response but shifted towards higher values of γ N 2 O 5 .At low values of γ N 2 O 5 the heterogeneous uptake of N 2 O 5 plays a negligible role in determining the NO x budget, thus in this regime, changes in γ N 2 O 5 have a small impact on burdens.At high values of γ N 2 O 5 , the rate limiting step for NO x loss switches from the heterogeneous uptake step (Reaction R4) to the production of NO 3 (Reaction R2) (which goes on to form N 2 O 5 ).Thus the model is again insensitive to the value of γ N 2 O 5 .It is only in the intermediate regime that the model shows sensitivity to the value of γ N 2 O 5 .The differing regional response is due to the aerosol loading being much higher in the northern extratropics, and temperatures are lower so the NO 3 production limitation occurs at lower values of γ N 2 O 5 .Globally the peak sensitivity for NO x is at a γ N 2 O 5 of 0.013, not far from the value of 0.02 found by Evans and Jacob (2005), whereas the peak sensitivities for O 3 and OH are at much higher γ N 2 O 5 of 0.16 and 0.2, respectively.Within the northern mid-latitudes the peak sensitivity lies at γ N 2 O 5 of 0.002 for NO x and at 0.03 for O 3 and OH.
The differing responses of the tropics, extra-tropics and globally leads to different conclusions about the importance of a precise description of γ N 2 O 5 .If we assume the true atmospheric value lies between 0.02 and 0.001 (as found by Evans andJacob, 2005 andBrown et al., 2009;Bertram et al., 2009) we find that the northern extra-tropics show a significant sensitivity to our choice of γ N 2 O 5 .Moving from 0.02 to 0.001 leads to NO x , O 3 and OH burdens in this region increasing by 29%, 7% and 8%, respectively.Thus, having a good definition of γ N 2 O 5 is important for defining the composition in this region.Globally the model is less sensitive with a change in γ N 2 O 5 from 0.02 to 0.001 leading to NO x , O 3 and OH burden changes of 11%, 3% and 4%, respectively.From our simulations we conclude that although an accurate defi-  nition of γ N 2 O 5 is significant for determining climate relevant parameters such as the global O 3 and OH burdens, it plays a much more significant role for the northern extra-tropics than the tropical regions.Thus for issues such as the long range transport of pollution, which is mostly a mid-latitude issue, conclusions drawn will be significantly impacted by the description of γ N 2 O 5 used.

Impact of aerosol loading on heterogeneous NO x loss
To a good approximation the rate of N 2 O 5 uptake is described by k= Aωγ 4 , where A is the aerosol surface area concentration, and ω is mean molecular speed.Therefore, the response of the model to a fractional perturbation in surface area, A, will be the same as a fractional perturbation in γ .Thus, our simulations provide insight into the impact of the choices of γ N 2 O 5 on model simulations where aerosol loading is changed (such as that between the pre-industrial and the present day).Figure 3 shows the percent change in burdens of NO x , O 3 and OH for a 10% reduction in aerosol surface area, (represented here by a 10% reduction in γ N 2 O 5 ) at different values of γ N 2 O 5 , calculated from the error function fit to the simulations from Fig. 2. Globally from our simulations we find very little variation in model sensitivity to aerosol perturbations in the range of γ N 2 O 5 1.0 down to 0.05, with sensitivity then dropping significantly below this value.At a γ N 2 O 5 of 0.001 as suggested by some studies, global O 3 and OH are significantly (an order of magnitude) less sensitive to perturbation in aerosol surface area than is the case at γ N 2 O 5 of 0.1.At a global scale, NO x appears to change by at most a factor 2 over this range in γ N 2 O 5 .This smaller sensitivity is due to the competing effect of increased sensitivity in the northern extra-tropics (by a factor of 3), and decreased sensitivity in the tropics and southern extra-tropics (an order of magnitude less).The northern extra-tropics again show a significantly different behaviour with peak sensitivity to an aerosol perturbation at around 0.05 for O 3 and OH.Thus for studies investigating the coupled impact of aerosol and chemistry changes (e.g., Bell et al., 2005;Lamarque et al., 2005;Liao et al., 2009), the conclusions are likely to be sensitive to the choice of γ N 2 O 5 .

Conclusions
Our model simulations show a non-linear tropospheric response to changes in γ N 2 O 5 with very small sensitivity at low values, and significant sensitivity at moderate values of γ N 2 O 5 (0.001-0.02).This response is regional due to differing aerosol loadings, temperature and photolysis.Within the likely range of γ N 2 O 5 (0.02 to 0.001) the northern extratropics show significant and enhanced sensitivity to the value of γ N 2 O 5 compared to the tropics and southern extra-tropics.
Models that use high values of γ N 2 O 5 (∼0.1) will overestimate the impact of changing aerosol loadings on composition through heterogeneous uptake.Thus a better understanding of the value of γ N 2 O 5 is needed to both understand current composition but also the combined impact of changing gasand aerosol-phase composition.

Fig. 2 .
Fig. 2. Impact of γ N 2 O 5 on mean annual burdens of NO x , O 3 , and OH.The curves are plotted using an error function fit to the points, which are taken from the mass weighted annual mean model diagnostics.Northern extra-tropics (90 • N-30 • N) are in red, tropics (30 • N-30 • S) are in green, southern extra-tropics (30 • S-90 • S) are in black, and global values are shown in blue.The vertical bar indicates the point of maximum gradient on the curve (the vertical black bar is not visible on the O 3 plot as it is overlain exactly by the green bar).

Fig. 3 .
Fig. 3. Impact of a 10% reduction in the product γ N 2 O 5 ×A (derived from the fits to the curves shown in Fig. 2) on mean annual burdens of NO x , O 3 , and OH for various γ N 2 O 5 values (shown on a log scale).Northern extra-tropics (90 • N-30 • N) are the red crosses, tropics (30 • N-30 • S) are green stars, southern extra-tropics (30 • S-90 • S) are black triangles, and global values are shown in blue circles.