Modeling the Formation, Degradation, and Spatiotemporal Distribution of 2-Nitrofluoranthene and 2-Nitropyrene in the Global Atmosphere.

Polycyclic aromatic hydrocarbons (PAHs) are common atmospheric pollutants and known to cause adverse health effects. Nitrated PAHs (NPAHs) are formed in combustion activities and by nitration of PAHs in the atmosphere and may be equally or more toxic, but their spatial and temporal distribution in the atmosphere is not well characterized. Using the global EMAC model with atmospheric chemistry and surface compartments coupled, we investigate the formation, abundance, and fate of two secondarily formed NPAHs, 2-nitrofluoranthene (2-NFLT) and 2-nitropyrene (2-NPYR). The default reactivity scenario, the model with the simplest interpretation of parameters from the literature, tends to overestimate both absolute concentrations and NPAH/PAH ratios at observational sites. Sensitivity scenarios indicate that NO2-dependent NPAH formation leads to better agreement between measured and predicted NPAH concentrations and that photodegradation is the most important loss process of 2-NFLT and 2-NPYR. The highest concentrations of 2-NFLT and 2-NPYR are found in regions with strong PAH emissions, but because of continued secondary formation from the PAH precursors, these two NPAHs are predicted to be spread across the globe.

Introduction

Human exposure to polluted air is associated with severe health effects: asthma, chronic obstructive pulmonary disease, lung cancer, cardiovascular diseases, and mortality.[1−4] Polluted air contains a complex mixture of organic and inorganic chemical species including polycyclic aromatic hydrocarbons (PAHs) and their photochemical degradation products, which exhibit mutagenicity and other hazardous properties.[5−10]

Nitrated PAHs (NPAHs) are a group of PAH derivatives substantially contributing to the toxicity of polluted air.[7,11−17] Many NPAHs are emitted together with PAHs from primary combustion sources; for example, 9-nitroanthracene, 3-nitrofluoranthene, and 1-nitropyrene are produced and emitted upon diesel combustion.[18−20] Other NPAHs are produced in the atmosphere by chemical transformation of PAHs. Specifically, 2-nitrofluoranthene (2-NFLT) and 2-nitropyrene (2-NPYR) are both formed by radical-initiated (OH or NO3) nitration of the parent PAH.[13,21−26] Unlike other isomers, 2-NFLT and 2-NPYR are not found in road and marine diesel emissions or only in minor amounts, for instance 2-NFLT/3-NFLT ≤ 0.05[20,27] and 2-NPYR/1-NPYR ≤ 0.01.[20,28] Nor are they emitted in coal or wood burning to any significant amount.[29−31]

Given the toxicity of NPAHs, there is a demand for assessing the exposure of humans and the environment. The biological effects of NPAHs are usually stronger than those of the parent PAHs. This has been documented for the mutagenicity of fluoranthene (FLT), pyrene (PYR), chrysene, and their nitrated derivatives[5,7,9,32,33] and for the developmental toxicity of phenanthrene, anthracene, PYR, and their nitrated derivatives.[17] The spatial and temporal distributions of 2-NFLT and 2-NPYR, however, have not yet been characterized. For these compounds, there is a lack of long-term air monitoring, and monitoring data are available only for few geographic locations.[16] As 2-NFLT and 2-NPYR are not directly emitted and their photochemical formation in air does not occur instantaneously, their distributions cannot be assumed to be identical to those of the parent PAHs. Because of long-range transport, both products[34−37] and their precursors[38,39] can be found in remote areas away from the source.[34,37,40]

Kinetic models have been used to explore the formation and loss of 2-NFLT and 2-NPYR,[41,42] while multimedia models have been used to quantify human exposure under steady-state conditions[33] or derive emission estimates for parent PAHs.[43] Regression models have been used to study 2-NPYR and 2-NFLT on local[44] and regional[45] scales. So far, however, atmospheric transport and chemistry have not yet been studied on the large scale, and the global atmospheric burden and geographical distribution of these secondarily formed NPAHs have not been addressed.

In this study, we implement the chemistry of 2-NFLT and 2-NPYR into a global model (EMAC-SVOC) which has previously been used to simulate the spatiotemporal distributions of FLT and PYR in the global atmosphere.[46] Nitration of the precursors FLT and PYR in the gas phase yields 2-NFLT and 2-NPYR, which then partition into the particle phase where they are susceptible to photodegradation. We present modeled near-surface concentrations and vertical column densities of 2-NFLT and 2-NPYR in comparison to observational measurement data from the literature, including a long-term monitoring site on the Noto Peninsula, Japan. Alongside the results using the simplest interpretation of literature input parameters (the default reactivity scenario), we present results from sensitivity scenarios to explore two uncertainties related to the chemical formation and loss of NPAHs in the atmosphere: the concentration of NO2 and the rate of photodegradation. Finally, we discuss the predicted spatial distributions of 2-NFLT and 2-NPYR column densities and near-surface concentrations.

Methods

Model Description and Setup

To simulate the chemistry and transport of 2-NFLT and 2-NPYR, we used the global ECHAM5/MESSy atmospheric chemistry model,[47,48] which includes the recently developed SVOC submodel to describe the environmental cycling of semivolatile organic compounds.[46] EMAC simulates gas-phase chemical reactions through the MECCA submodel[49] and calculates photolysis rates through the JVAL submodel.[47] EMAC-SVOC simulates diffusive air–surface exchange through coupled surface compartments and was recently evaluated for PAHs with a wide range of physicochemical properties.[46] The oxidants OH, O3, NO2, and NO3 are tracers formed and consumed in MECCA. The scavenging of gases and aerosols by clouds and rain is calculated by the SCAV submodel,[50] while their dry removal mechanism is calculated by DRYDEP.[51] Aerosol microphysics (external/internal mixing, nucleation, condensation, and coagulation) are treated by the GMXe submodel.[52] The primary aerosol components included are sulfate, black carbon (BC), organic matter (OM), sea salt, and mineral dust. Each aerosol species is described as seven log-normal size distributions, of which four modes (nucleation, Aitken, accumulation, and coarse) are soluble and three modes (Aitken, accumulation, and coarse) are insoluble. The nucleation hydrophilic mode is only used to describe the sulfate particles.[52] We neglect the treatment of the nitrate aerosol but include that of ammonium; consequently, the model assumes that all sulfate is in the form of ammonium sulfate.[52] The effect of ammonium nitrate aerosols on NPAH tracers is expected to be negligible as NPAH formation and photodegradation are independent of nitrate concentrations and even pH. The contribution of adsorption to NH4NO3 to gas/particle partitioning is minor as adsorption in total is minor compared to absorption in OM.[53] In the model aerosol, the NPAH is concentrated in the submicrometer size fraction (about 80% are associated with PM1), in accordance with observations.[54−57]

The model was run on a T42 spectral grid resolution (corresponding to a Gaussian grid of approximately 2.8 × 2.8° lat-lon) and 19 vertical hybrid pressure levels extending to 10 h Pa. The simulations were performed for the years 2005–2008, including a 1 year spin-up period, and were nudged toward the European Centre for Medium-Range Weather Forecast (ECMWF) reanalysis data.[58]

The formation of 2-NFLT and 2-NPYR (from their parent PAHs) as well as their loss was implemented in this model as detailed below.

Formation of 2-NFLT and 2-NPYR

The dominant source of 2-NFLT and 2-NPYR is nitration of the respective gaseous PAH precursors FLT and PYR. Heterogeneous reactions of PYR and NO2 on minerals form 1-NPYR only,[59−61] while from FLT, no 2-NFLT is formed under ambient NO2 concentrations, unlike under unrealistically high NO2.[62] It is now understood that the heterogeneous reactions of particle-bound FLT and PYR with ambient levels of NO and NO2/O3 and under realistic atmospheric conditions (e.g., relative humidity) do not allow for or only result in negligible formation of 2-NFLT and 2-NPYR.[62−64] Consequently, 2-NFLT and 2-NPYR in the atmosphere are mainly attributable to their gas-phase free-radical formation (described below).

Gas-phase formation of 2-NFLT and 2-NPYR is included in the model. Table lists the reactions, and Table lists the rate coefficients and yields used. This formation proceeds via a two-step reaction (Figure S1).[65] In the first step, a hydroxyl (OH) or nitrate (NO3) radical reacts with the PAH to form a PAH-radical adduct. This intermediate may react with either NO2 to form NPAH or O2 to form other oxygenated products.[13,65] Note that 2-NFLT formation is initiated by either OH or NO3, whereas 2-NPYR formation is initiated by OH attack only. Atkinson et al. empirically determined rate coefficients for the addition reactions and calculated the overall yields of 2-NFLT and 2-NPYR at NO concentrations (≈1–10 ppm) much higher than typical ambient conditions.[66] In addition, Brubaker and Hites experimentally determined the rate coefficient for the reaction of FLT with OH.[67] These rate coefficients and yields are used in the model as the default reactivity scenario in the model (Table , default reactivity scenario).

Table 1

Overview of Chemical Formation and Loss Reactions in the Model for 2-NFLT and 2-NPYR

reaction	rate expression
FLT(g) + OH(g) → 2-NFLT(g)	Y2-NFLT,OH × kFLT+OH × [OH]g × [FLT]g
PYR(g) + OH(g) → 2-NPYR(g)	Y2-NPYR,OH × kPYR+OH × [OH]g × [PYR]g
FLT(g) + NO3(g) → 2-NFLT(g)	Y2-NFLT,NO3 × kFLT+NO3 × [NO3]g × [FLT]g
2-NFLT(p) + hν → products	α × JNO2 × [2-NFLT]p
2-NPYR(p) + hν → products	α × JNO2 × [2-NPYR]p

Table 2

Overview of Chemical Formation and Loss Reactions in the Default Reactivity Scenario for 2-NFLT and 2-NPYR and Three Sensitivity Scenarios (α = 0.005, NO2-Dependent Yield with α = 0.05 and NO2-Dependent Yield with α = 0.005)a

parameter	default α = 0.005	α = 0.05	NO2-dependent yield/α = 0.05	NO2-dependent yield/α = 0.005	units
kFLT+OH	1.1 × 10–11b	1.1 × 10–11	1.1 × 10–11	1.1 × 10–11	cm3 s–1
kPYR+OH	5.0 × 10–11c	5.0 × 10–11	5.0 × 10–11	5.0 × 10–11	cm3 s–1
kNFLT+NO3	0.51 × [NO2] × 10–27c	0.51 × [NO2] × 10–27	0.51 × [NO2] × 10–27	0.51 × [NO2] × 10–27	cm3 s–1
Y2-NFLT,OH	0.03	0.03	0.03 × Ω	0.03 × Ω
Y2-NPYR,OH	0.005	0.005	0.005 × Ω	0.005 × Ω
Y2-NFLT,NO3	0.24	0.24	0.24 × Ω	0.24 × Ω
α	0.05d	0.005	0.05	0.005

kFLT+OH, kPYR+OH, and kFLT+NO are the rate coefficients for each reaction, and Y2-NFLT,OH, Y2-NPYR,OH, and Y2-NFLT,NO are the NPAH yields. In the NO2-dependent simulation, a scaling factor is used, Ω = (kNO/kO) ([NO2]g/[O2]g)/(1 + (kNO/kO) ([NO2]g/[O2]g)) with kNO/kO = 1 × 107. α is a factor to scale NPAH photodegradation relative to the NO2 photolysis rate (Figure S2).

Brubaker and Hites, 1998.[67]

Atkinson et al., 1990.[66]

Fan et al., 1996.[71]

As previously noted, a key step in the formation mechanism involves competition between NO2 and O2 for an intermediate (Figure S1). Therefore, under ambient and particularly low NO conditions, one would expect the yield of these compounds to decrease and to be dependent on NO2 concentration. The yield determined by Atkinson et al. likely represents an upper limit[66]; therefore, using the default reactivity scenario in the model might overestimate the conversion of the PAH to the NPAH under low NO conditions. In order to explore this uncertainty in more detail, an alternative formation scenario, in which the yields of 2-NFLT and 2-NPYR are dependent on NO2 concentration, is tested (Table , NO2-dependent yield). Relative rate coefficients for the reactions between the intermediate and either O2 or NO2 are used to create an expression for the yield as a function of NO2 concentration (detailed in the Supporting Information, Section S1, and Table ; the explicit dependence of yield on NO2 can be seen in Figure S2).

Gas-Particle Partitioning and Deposition

The mass fractions of the PAH and NPAH in the gas and particle phases are set to the predicted thermodynamic gas–particle partitioning equilibrium at each model time step. These equilibrium predictions are based on an implementation of the ppLFER approach,[53,68] for which detailed information is available in the Supporting Information. The physicochemical parameters of each PAH and NPAH species used in the model are given in Table S1.

In the gas phase, atmospheric sources of semivolatile organics (besides chemical formation) are primary emissions and revolatilization from ocean and terrestrial surfaces, while their sinks are wet scavenging and dry deposition.[46] In the particle phase, the dry deposition and wet deposition of PAHs and NPAHs are treated the same as that of other particle-phase species.

Chemical Losses of 2-NFLT and 2-NPYR

The most important loss pathway of 2-NFLT and 2-NPYR is sunlight-initiated photodegradation in the particle phase.[69] The exact mechanism for these two particular NPAHs is unknown, although 1-NPYR has been shown to have a complex degradation mechanism.[70] Photolysis in the gas phase was never observed.[23] Fan et al. quantified the photodegradation rates of 2-NFLT and 2-NPYR relative to the photolysis rate of NO2 (JNO).[71] This was parameterized using a scaling factor α, which was determined for several types of soot particles only. For wood smoke particles, α = 0.05 was found, but on diesel soot particles, 2-NFLT and 2-NPYR were slightly less reactive, with α = 0.025 and α = 0.04, respectively.

The sole chemical sink of the atmospheric NPAH implemented in this model is photolytic degradation, which was included only for 2-NFLT and 2-NPYR in the particle phase. For the default reactivity scenario, we adopt the scaling factor from the wood smoke aerosol (α = 0.05) for 2-NFLT or 2-NPYR (Table ). JNO is calculated for each grid cell based on available sunlight. The constant scaling factor is applied for all aerosols, despite the fact that particulate matter (PM) composition and color strongly influence the photochemical degradation process.[70,72−75] Photolysis with different efficiencies might be expected in aged, long-range transported aerosols and in natural aerosols (sea salt, mineral dust).

We explore this uncertainty using an alternative reactivity scenario with the loss rate due to photodegradation reduced by a factor of 10 (i.e., α = 0.005, Table ) and also simulated with photolysis switched off (α = 0).

Another possible NPAH loss reaction is the homogeneous reaction with the OH radical. Second-order rate coefficients have not been determined experimentally, but estimates are available, that is, 4.93 × 10–12 and 6.25 × 10–12 molecules cm–3 s–1 for 2-NFLT and 2-NPYR, respectively.[76] A sensitivity run including these reactions showed only a small effect on simulation results (Figure S3). Therefore, the homogeneous reaction of NPAHs with OH is not further included in simulations.

PAH Emissions

FLT and PYR emission data of the year 2008 were used in the model simulation.[77] The total global emissions of FLT and PYR were 2.8366 × 107 and 2.1827 × 107 kg yr –1, respectively. These emissions comprise fossil fuel (coal, oil, and gas), biomass, and waste combustion for energy production, industry, transportation (including ships and aircraft), and agricultural and residential sources. Emissions from industrial processes and open fires are also included in the inventory. The emissions were projected from a country-level to a global grid (with a resolution of 0.1 × 0.1°) and represented as annual sums. For the simulations, we applied monthly variation of PAH emissions following the temporal profile of emissions of BC from the Hemispheric Transport of Air Pollutants (HTAP) v2.2 data set.[78] PAHs are largely coemitted with BC.[77,79] Emissions of other gas and aerosol species were the same as those in Octaviani et al. (2019).[46]

Comparison between Observational Measurements and Simulation

The model output was compared with measurement data of 2-NFLT, 2-NPYR, FLT, and PYR at different locations, spanning the years 2000–2017 (see Table S2 for details). Ambient air concentrations of PAHs and NPAHs in North America, Europe, and Japan have decreased strongly during the 1990s because of the introduction of emission control measures.[43,80−82] Hence, we used only measurements made since the year 2000 for comparison to the simulation/emission data in this study (2006–2008).

Note that the measurement data comprise total (gas phase and particle phase) as well as particle phase-only measurements. Similarly, not all studies analyzed the same size fractions of particulate matter, and in several studies, 2-NFLT data included 3-nitrofluoranthene (3-NFLT) because of chromatographic coelution.[16] However, 3-NFLT abundance in ambient air is very low and can be considered negligible with respect to 2-NFLT.[20] Nevertheless, these inconsistencies may contribute to discrepancies between the model and observations.

For comparison with observational data, the simulated near-surface concentrations of the NPAH and PAH were bilinearly interpolated between grid cells and given as a monthly mean (see the Supporting Information, Section S4). This was done for both the particle-phase only and total near-surface concentrations. Furthermore, observational sites are categorized as either rural/remote, that is, sites away from the source where sampled air is aged or from unpolluted vicinity, or urban, that is, polluted sampling sites close to the source such as cities (Table S2). Because the coarse resolution of the large-scale model is unable to resolve the subgrid level heterogeneity induced by local sources, the model systematically underestimates the measured concentration of both 2-NFLT and 2-NPYR at the sites classified as urban (Figures S2 and S3). Hence, we focus on the sites classified as rural in the further discussion. The distinction between rural and urban sites was made by analyzing keywords from the original publications. Sites described by keywords such as “city”, “suburban”, and “highway” are classified as urban, while those described by “rural”, “background”, or “remote” are denoted as rural. A full list of the keywords associated with each site and their subsequent classification are shown in the Supporting Information (Table S2). Comparisons between simulated and measured near-surface concentrations at rural and urban sites for 2-NFLT and 2-NPYR are shown in the Supporting Information (Figures S4 and S5).

The correlation coefficient (R) and modified normalized mean bias (MNMB) are used to evaluate model performance. The MNMB has been previously used to evaluate predicted atmospheric concentrations that vary across orders of magnitude.[83−85] The MNMB is calculated in the following waywhere N is the number of simulation–observation pairs and f and o are the numerical values of simulation and observation, respectively. The MNMB may also be expressed as a percentage and can range between −200% and +200%.

Results and Discussion

Comparison with Field Observations across the Globe

Default Reactivity Scenario

Figure shows a comparison of the simulated and measured near-surface concentrations for 2-NFLT and 2-NPYR for the default reactivity scenario and three alternative scenarios with different chemical formation and degradation parameters. The measured concentrations of 2-NFLT and 2-NPYR are spread across 5 and 3 orders of magnitude, respectively (6 × 10–2 to 4 × 103 pg m–3 for 2-NFLT and 1 × 10–1 to 7 × 101 pg m–3 for 2-NPYR). In the default reactivity scenario (α = 0.05, indicated by blue points), we find overestimation and underestimation of 2-NFLT contained within 1 order of magnitude. The MNMB is +10%, thus indicating that the concentration of 2-NFLT is slightly overestimated by the default reactivity scenario (Table ). The measured 2-NFLT concentrations below 10 pg m–3 tend to be overestimated by up to 1 order of magnitude, while for concentrations higher than 10 pg m–3, the model results are scattered above and below the 1:1 line (Figure a).

Figure 1

Comparison between simulated and measured near-surface concentrations [pg m–3] at rural sites of (a) 2-NFLT and (b) 2-NPYR. The dashed lines marked 1:1, 100:1, and 1:100 represent where the measured concentrations are predicted, underestimated by a factor of 100, and overestimated by a factor of 100, respectively.

Table 3

Correlation Coefficient (R; * = Significant at P < 0.05) and MNMB to Evaluate the Comparison between the Different Simulations and Measurementa

	2-NFLT(p)		2-NFLT(p)/FLT(p)		2-NPYR(p)		2-NPYR(p)/PYR(p
scenario	R	MNMB (%)	R	MNMB (%)	R	MNMB (%)	R	MNMB (%)
α = 0.05	0.10	10	–0.10	+139	–0.09	–27	0.50	+117
α = 0.005	0.16	+89	–0.10	+168	–0.12	+51	0.71*	+170
NO2-dependent yield/α = 0.05	–0.03	–115	–0.07	–2	–0.11	–173	0.19	–112
NO2-dependent yield/α = 0.005	–0.01	–66	–0.07	+69	–0.11	–138	0.78*	–29

For 2-NFLT (N = 48), 2-NFLT/FLT (N = 24), 2-NPYR (N = 18), and 2-NPYR/PYR (N = 10).

Importantly, FLT, the precursor to 2-NFLT, tends to be underestimated at these rural sites (Figure S6), suggesting that the true model overestimation of 2-NFLT may be even greater. In order to normalize for the effects of the parent PAH concentration, the ratio of NPAH/PAH near-surface concentrations is calculated. A comparison between the simulated and observed ratios of 2-NFLT to FLT concentration is shown (Figure ). The default chemistry scenario systematically overestimates the 2-NFLT-to-FLT ratio and has an MNMB equal to +139% (Table ).

Figure 2

Comparison between the simulated and measured ratios of NPAH/PAH near-surface concentrations at rural sites of (a) 2-NFLT/FLT and (b) 2-NPYR/PYR. The dashed lines marked 1:1, 100:1, and 1:100 represent where the measured concentrations are predicted, underestimated by a factor of 100, and overestimated by a factor of 100, respectively.

Predictions of 2-NPYR using the default chemistry scenario are scattered within 1 order of magnitude above and below the 1:1 line with a small negative bias (MNMB = −27%). However, as in the case of 2-NFLT, the ratio of 2-NPYR to PYR is overestimated (MNMB = +117%).

Measurement data for observational sites where both the NPAH and the precursor PAH have been measured indicate that the NPAH/PAH ratio varies across 3 orders of magnitude (Figure ). A spatially homogeneous ratio would indicate that the concentration of the precursor PAH is strongly correlated with the NPAH product, whereas both the measurement data and our simulation results clearly demonstrate that these ratios are highly variable. Apart from the distribution of PAH emissions, NPAH variability is determined by the availability of oxidants, sunlight, NO, and aerosol mass concentrations (with the latter influencing the extent to which the PAH and NPAH partition between the gas and particle phases).

By dividing the global atmospheric burdens through the overall loss rates related to chemical and physical sinks, we obtain global mean atmospheric residence times of 2.6 h for 2-NFLT and 2.9 h for 2-NPYR (default reactivity scenario; Table ). Dividing the global atmospheric burdens by the photodegradation loss rates gives photolytic lifetimes, for which we obtain global means of 2.8 h for 2-NFLT and 3.2 h for 2-NPYR. This comparison shows that the atmospheric fate of 2-NFLT and 2-NPYR under these reactivity scenarios is mostly driven by photodegradation, with significantly lower contribution from dry and wet deposition.

Table 5

Total Emission Fluxes of Precursors (FLT and PYR)[77] and Global Mean Parameters for the Atmospheric Cycling of 2-NFLT and 2-NPYR for the Years 2006–2008 (over Land/over Ocean) Using the Default Reactivity Scenario and the Three Additional Sensitivity Scenarios

species	scenario	precursor emission flux (×103 kg yr–1)	global atmospheric burden (kg)	near-surface concentration (pg m–3)	atmospheric residence time (h)	photolytic lifetime (h)
2-NFLT	α = 0.05	FLT: 28366 (27255/1111)	929	2.93 (8.24/0.84)	2.59	2.81
	α = 0.005		4170	10.9	17.2	27.9
	NO2-dependent yield/α = 0.05		86.2	0.5	3.01	3.26
	NO2-dependent yield/α = 0.005		288	1.04	18.7	31.3
2-NPYR	α = 0.05 default	PYR: 21827 (20533/1294)	100.1	0.43 (1.39/0.06)	2.87	3.15
	α = 0.005		510	1.59	18.1	30.8
	NO2-dependent yield/α = 0.05		2.08	0.0173	2.57	2.8
	NO2-dependent yield/α = 0.005		10.7	0.048	17.6	29.2

Given its importance, the photodegradation rate is rather poorly constrained and may strongly contribute to model uncertainty. Kinetic data describing the effects of particle composition, and possibly phase state, on heterogeneous photolysis would be needed but are not available. Quantum mechanical calculations have provided a detailed understanding of the photochemistry of 1-nitropyrene,[86,87] and similar approaches would be valuable for 2-NFLT and 2-NPYR.

Alternative Scenarios

To explore the overestimation bias and the lack of correlation between the simulated predictions of the default reactivity scenario and measured near-surface concentrations of 2-NFLT and 2-NPYR (Figure ) and NPAH/PAH ratios (Figure ), we present the results from three alternative reactivity scenarios. These include a NO2-dependent NPAH formation scenario which should lead to a higher sensitivity of the model to local conditions, a scenario with a reduced photochemical loss of NPAH, and a combined NO2-dependent/reduced photochemical loss scenario.

The NO2-dependent reactivity scenarios account for reduced NPAH yields under low NOx conditions. Their implementation leads to decreased atmospheric NPAH column densities in less polluted regions, while in regions with elevated NO2 concentrations, the formation is similar to that in the default scenario (Figure S7). The global atmospheric burden of 2-NFLT decreases by a factor of 11, and the global atmospheric burden of 2-NPYR decreases by a factor of 48 (Table ). Correspondingly, observed 2-NFLT and 2-NPYR concentrations are underestimated (MNMB = −115% and −173%, respectively, for α = 0; Table ). In the source regions (Figure ; almost all these rural sites are located in source regions), NPAH near-ground concentrations are underpredicted by up to 3 orders of magnitude, while at the remote site (Figure ), NPAH concentrations are underpredicted by up to 1 order of magnitude. With the NO2-dependent scenario, the MNMB of the 2-NFLT-to-FLT ratio is only −2%, which is thus much improved compared to that of the default scenario. For the 2-NPYR-to-PYR ratio, the MNMB does not improve significantly (−112%).

Figure 3

Comparison between simulation and observations for the Noto Peninsula in Japan (January 2006 to December 2007) for (a) 2-NFLT and (b) 2-NPYR particle-phase concentrations [pg m–3]. Measurement data are from Tang et al. (2014)[88] and presented as monthly mean concentrations (solid red line). Error bars in the measurement data are the minimum and maximum weekly values for each month. The upper and lower quartiles of simulated concentrations each month are bounded by the shaded region. The modified normalized mean bias (MNMB) and Pearson correlation coefficient (R) are shown for the default simulation (α = 0.05, blue line).

The two reduced photodegradation scenarios account for the large uncertainty connected with this model parameter. Slowing the rate of loss due to photodegradation by a factor of 10 leads to an increase in 2-NFLT and 2-NPYR concentrations at all observation sites (Figure ). Relative to the default scenario (α = 0.05), this α = 0.005 scenario demonstrates even more overestimation by the model: the MNMBs are +89, +51, +168, and +170% for 2-NFLT, 2-NPYR, 2-NFLT/FLT, and 2-NPYR/PYR, respectively (Table , Figures and 2). The only metric for which reducing the rate of photolysis resulted in a statistically significant (P < 0.05) improvement in correlation is 2-NPYR/PYR (R = 0.71).

Note that fully neglecting photolytic degradation (α = 0, no photolysis) would lead to massive overestimation of atmospheric column densities of 2-NFLT and 2-NPYR by 2 orders of magnitude, as illustrated by a sensitivity test. Spatially, this is reflected as higher concentrations in the continental plumes from northern midlatitude and tropical source regions (Figure S7). Regions far from the source (e.g., Antarctica) tend to be the most sensitive to effects of photodegradation and have a column density of more than 3 (2-NFLT) and 4 (2-NPYR) orders of magnitude greater compared to that of the default reactivity scenario (Figure S7). In contrast, in source regions (e.g., Europe) increases are only up to 1 order of magnitude. These results demonstrate that the rate of photodegradation becomes more influential for the atmospheric column concentrations of 2-NFLT and 2-NPYR the farther air masses are transported.

Assuming both a NO2-dependent formation of the NPAH and a low estimate for NPAH photodegradation (α = 0.005), the global atmospheric burden of 2-NFLT decreases by a factor of 3 and the global atmospheric burden of 2-NPYR decreases by a factor of 9 (Table ), leading to underestimation of 2-NFLT and 2-NPYR concentrations (MNNB = −66 and −138%, respectively; Figure ). However, among the four scenarios investigated, the combined NO2-dependent/low photodegradation scenario leads to the lowest discrepancies of 2-NFLT/FLT (MNMB = −2 and +69% for α = 0.05 and 0.005, respectively) and 2-NPYR/PYR ratios (MNMB = −112 and −29% for α = 0.05 and 0.005, respectively).

In conclusion, a NO2-dependent scenario is able to reduce some of the model overestimation that occurs for 2-NFLT in less polluted areas (discussed in Section and seen in Figure a) and comes closest to observed NPAH/PAH ratios. Further exploration of the 2-NFLT and 2-NPYR formation kinetics at low NO concentrations with laboratory experiments may ultimately help to improve model performance.

Comparison with Observations at a Remote-Background Site with Long-Term Monitoring

The Wajima Air Monitoring Station on the Noto Peninsula, Japan (37°23′ N, 136°54′ E; Tang et al. 2014) represents one of very few long-term monitoring data sets of 2-NFLT and 2-NPYR and the only one at a remote-background location.[88] Remote-background locations are less likely to be influenced by strong local concentrations and are therefore particularly valuable for assessing the performance of global models. Monthly averages of 2-NFLT and 2-NPYR concentrations (Tang et al.) between January 2006 and December 2007 are shown in Figure and compared with simulation predictions.

In the default reactivity scenario (α = 0.05), the model tends to overestimate the concentration of 2-NFLT and 2-NPYR at this site (MNMB = +87 and +96%, respectively); however, simulated particle-phase concentrations of 2-NFLT and 2-NPYR agree with measurements within 1 order of magnitude (Figure ). For 2-NFLT, this overestimation is most prevalent in summer. The winter maximum of NPAHs (October to April) was attributed to transport from Northeast Asia, during the heating period in Northern China.[88] Measured 2-NFLT concentrations exhibit a clearer seasonal signal than 2-NPYR; the model thus captures this seasonal variability of 2-NFLT (Figure a, blue line, R = 0.52, p < 0.05) slightly better than that for 2-NPYR (Figure b, blue line, R = 0.50, p < 0.05). Throughout the 2-year period, simulated concentrations of 2-NFLT range between 2 and 8 pg m–3. In agreement with measurements, the simulated concentrations of 2-NPYR are around a factor of 10 lower than those of 2-NFLT (≈0.2 to 1.4 pg m–3).

The MNMB for the simulated ratio of concentrations is higher than that for absolute concentrations: 2-NFLT/FLT (+169%) and 2-NPYR/PYR (+174%). In contrast to the absolute concentration, 2-NFLT/FLT and 2-NPYR/PYR ratios show maxima in summer (Figure ). The measured ratios, calculated from the data of Tang et al.,[88] also exhibit maxima in summer but are less extreme. For example, the mean measured 2-NFLT/FLT ratios for January and July are 0.009 and 0.032, respectively. The model produces a similar pattern, with the simulated mean 2-NFLT/FLT ratios being 0.20 and 3.49 in January and July, respectively. Two processes in the model may explain the summer maxima of these ratios: (1) the increased availability of oxidants (OH and NO3) and (2) higher temperatures causing a higher proportion of FLT and PYR to be in the gas phase and therefore convert to 2-NFLT and 2-NPYR.

Figure 4

Comparison between simulation and observations for the Noto Peninsula in Japan (January 2006 to December 2007) for (a) 2-NFLT/FLT and (b) 2-NPYR/PYR ratio of particle-phase concentrations. Measurement data are from Tang et al. (2014)[88] and presented as monthly mean concentrations (solid red line). The modified normalized mean bias (MNMB) and Pearson correlation coefficient (R) are shown for the default simulation (α = 0.05, blue line).

PAH partitioning is crucial for the ability of the model to predict NPAH concentrations because NPAHs are formed in the gas phase but are primarily lost from the particle phase by photodegradation. Therefore, the transfer of mass between phases may be an important process controlling the atmospheric fate of the NPAH. The model predicts the near-surface gas-phase concentration of 2-NFLT and 2-NPYR to be lower compared to that in the particle phase (corresponding to a particulate mass fraction generally over 80%, even in summer). This is consistent with previous studies which have found these NPAHs almost exclusively in the particle phase.[53,89−91] The model predicts that on average over the 2-year period, only small fractions of the total mass of FLT and PYR in air are present in the particle phase (7 and 14%, respectively); in contrast, 2-NFLT and 2-NPYR are mainly associated with particles (91 and 90%, respectively). Also, the simulated parent PAH gas–particle partitioning is consistent with previous findings at other sites.[67,90,91]

The NO2-dependent formation scenario, with default photodegradation (α = 0.05), underestimates absolute 2-NFLT and 2-NFLT concentrations at the Noto Peninsula site (Table , brown dotted line in Figure a) but better predicts 2-NFLT/FLT and 2-NPYR/PYR ratios compared to the default chemistry scenario (Table , brown dotted line in Figure ). The scenario with reduced photodegradation (α = 0.005, green dotted lines in Figures and 4) leads to more severe overestimation of absolute NPAH concentrations and NPAH/PAH compared to default chemistry scenario. Model performance improves when the NO2-dependent formation scenario is used with reduced photodegradation (α = 0.005, yellow dotted line in Figures and 4). A slight underestimation of absolute NPAH concentrations is still observed, but NPAH/PAH ratios are better predicted compared to the default scenario.

Table 4

Correlation Coefficient (R; * = Significant at P < 0.05) and MNMB to Evaluate the Comparison between the Different Simulations and Measurement at the Noto Sitea

	2-NFLT(p)		2-NFLT(p)/FLT(p)		2-NPYR(p)		2-NPYR(p)/PYR(p)
scenario	R	MNMB (%)	R	MNMB (%)	R	MNMB (%)	R	MNMB (%)
α = 0.05	0.52*	+87	0.63*	+169	0.50*	+96	0.36	+174
α = 0.005	0.51*	+169	0.48*	+192	0.51*	+171	0.25	+193
NO2-dependent yield/α = 0.05	0.21	–167	0.69*	–65	–0.02	–165	0.40	–48
NO2-dependent yield/α = 0.005	0.30	–78	0.53*	+66	–0.01	–103	0.27	+56

For 2-NFLT, 2-NFLT/FLT, 2-NPYR (N = 24), and 2-NPYR/PYR (N = 23).

Global Atmospheric Burden, Near-Surface Concentrations, and Column Densities

Figure a shows the model-calculated near-surface concentrations of 2-NFLT and 2-NPYR averaged over 3 years, 2006–2008. They exhibit strong spatial variability, spanning over 6 orders of magnitude from near-source to remote marine environments and polar regions (from 10–3 to 103 pg m–3). The mean near-surface concentrations of 2-NFLT and 2-NPYR in the Arctic are ≈1 and ≈0.1 pg m–3, respectively, and a factor of 3–5 higher in low-latitude remote regions such as the Amazon and Tibet. The mean concentrations of 2-NFLT and 2-NPYR over the midlatitude open North Atlantic are ≈10 and ≈1 pg m–3, respectively, and 2 orders of magnitude lower over the midlatitude open North Pacific. The highest average concentration levels within a model grid cell range up to ∼1000 pg m–3 for 2-NFLT and ∼10 pg m–3 for 2-NPYR and are located in PAH source regions: Europe, Western Russia, Central Africa, and East and Southeast Asia. The simulated global mean concentrations of 2-NFLT and 2-NPYR at the near-surface level are 2.93 and 0.43 pg m–3, respectively (Table ). Between individual years, the global mean near-surface concentrations differed by less than 20% for 2-NFLT and less than 15% for 2-NPYR.

Figure 5

(a) Atmospheric concentrations at the near-surface level [pg m–3] of 2-NFLT and 2-NPYR and (b) column densities [kg m–2], averaged over 2006–2008, using the default reactivity scenario.

The column densities, that is, the vertically integrated concentration per unit area, of 2-NFLT and 2-NPYR exhibit a similar geographic distribution to the near-surface concentrations (Figure b) and range across at least 6 orders of magnitude, from 10–16 to 10–10 kg m–2. The highest column densities are located in PAH source regions, ∼10–10 kg m–2 for 2-NFLT and ∼10–11 kg m–2 for 2-NPYR. The global atmospheric burdens, that is, the total masses of 2-NFLT and 2-NPYR stored in the global atmosphere, are 929 and 100 kg, respectively (averaged over 2006–2008, Table ). These correspond to 0.7 and 0.2%, respectively, of the parent PAHs’ global atmospheric burden. The global atmospheric burden differed between individual years by less than 8% for 2-NFLT and less than 9% for 2-NPYR. In spite of the low mean atmospheric residence times (2.6 and 2.9 h), 2-NFLT and 2-NPYR are spread across the globe. This is due to continuous formation of the NPAHs from photochemical reactions in the atmosphere that take place as the PAH precursors are transported away from source regions (Figure ). This can be highlighted by the lifetimes of the precursors toward their main chemical sink, homogeneous reaction with the OH radical, of approximately 24 h and 5 h for FLT and PYR, respectively (using [OH] = 1.16 × 106 cm–3 [65] with kFLT-OH and kPYR-OH in Table ), causing the parent compounds to be transported farther than their nitrated degradation products.

Figure shows the ratio of near-surface concentrations for (a) 2-NFLT to FLT and (b) 2-NPYR to PYR, averaged over the years 2006–2008. The spatial distributions of these ratios vary across the globe, ranging across at least 3 orders of magnitude. The ratio of 2-NFLT to FLT tends to be around 0.1–1% over continental source regions, whereas the highest ratio occurs in plumes being transported away from sources. Particularly high ratios occur over the ocean and in some cases exceed 10%. A similar tendency was seen for the 2-NPYR-to-PYR ratio, which is generally around 0.1% in near-source regions but increases up to 1% in plumes being transported to more remote regions.

Figure 6

Ratio of NPAH to parent PAH near-surface concentrations, (a) 2-NFLT/FLT and (b) 2-NPYR/PYR, averaged over 2006–2008.

In conclusion, our model predicts the highest concentrations of 2-NFLT and 2-NPYR to be found in regions with high PAH emissions, but because of formation along long-range atmospheric transport, the two NPAHs are distributed worldwide.