Large contribution of biomass burning emissions to ozone throughout the global remote troposphere.

Ozone is the third most important anthropogenic greenhouse gas after carbon dioxide and methane but has a larger uncertainty in its radiative forcing, in part because of uncertainty in the source characteristics of ozone precursors, nitrogen oxides, and volatile organic carbon that directly affect ozone formation chemistry. Tropospheric ozone also negatively affects human and ecosystem health. Biomass burning (BB) and urban emissions are significant but uncertain sources of ozone precursors. Here, we report global-scale, in situ airborne measurements of ozone and precursor source tracers from the NASA Atmospheric Tomography mission. Measurements from the remote troposphere showed that tropospheric ozone is regularly enhanced above background in polluted air masses in all regions of the globe. Ozone enhancements in air with high BB and urban emission tracers (2.1 to 23.8 ppbv [parts per billion by volume]) were generally similar to those in BB-influenced air (2.2 to 21.0 ppbv) but larger than those in urban-influenced air (-7.7 to 6.9 ppbv). Ozone attributed to BB was 2 to 10 times higher than that from urban sources in the Southern Hemisphere and the tropical Atlantic and roughly equal to that from urban sources in the Northern Hemisphere and the tropical Pacific. Three independent global chemical transport models systematically underpredict the observed influence of BB on tropospheric ozone. Potential reasons include uncertainties in modeled BB injection heights and emission inventories, export efficiency of BB emissions to the free troposphere, and chemical mechanisms of ozone production in smoke. Accurately accounting for intermittent but large and widespread BB emissions is required to understand the global tropospheric ozone burden.

Tropospheric ozone (O3) has been the focus of decades of scientific research due to its central role in atmospheric chemistry (1), its adverse impact on human and ecosystem health (2, 3), and its role as a climate forcer (4, 5). Despite this focus, there remains considerable uncertainty in tropospheric O3 production pathways, precursor sources, and long-term trends. Sources of tropospheric O3 include downward transport from the stratosphere and photochemical production from a complex set of coupled reactions between nitrogen oxides (NOx) and volatile organic compounds (VOCs), each of which is in turn emitted from both anthropogenic and natural sources (1, 6). The contribution of fossil fuel combustion to tropospheric O3 has recently declined in the United States and in Europe, proportionally increasing the contribution from natural sources (7–10). However, the spatial distribution of anthropogenic O3 precursor emissions have shifted to lower latitudes (11, 12), where they are still increasing (13). Additionally, globally averaged tropospheric O3 has increased over the past five decades (14, 15). Understanding the sources of tropospheric O3 is thus essential to explain this trend and to inform the development of effective mitigation strategies from regional to hemispheric scales.

Biomass burning (BB) is an important source of O3 precursors (16–19). A recent study based on observed O3 to carbon monoxide (CO) enhancements in smoke plumes attributed 3.5% of the global tropospheric chemical O3 production to BB emissions (19). Other studies have accounted for the numerous production and destruction pathways of O3 in the troposphere using global chemical transport models (CTMs) to estimate the global budget of O3 (20, 21). However, few studies separately quantify the contributions of fossil fuel combustion and BB emissions to global tropospheric O3 (22). Global inventories attribute five times more NOx (23, 24) but roughly equal VOC emissions (17, 25) to fossil fuel combustion (hereafter referred to as urban sources) compared with BB. However, precursor emissions do not necessarily determine tropospheric O3 production close to the sources because of the nonlinearity of O3 formation chemistry (26, 27). Additionally, global CTMs do not always agree on the tropospheric O3 burden, suggesting possible deficiencies with emission inventories of O3 precursors and/or an incomplete representation of O3 chemistry (21, 28–30), although a recent model intercomparison study showed that the model ensemble reproduced well the salient spatial, seasonal, and decadal variability and trends of tropospheric O3 (31).

Large-scale in situ observational constraints commensurate with the grid resolution of current global CTMs are rare. Instead, modeling studies often rely on ozonesonde-derived climatologies and satellite-based remote sensing observations to constrain tropospheric O3 distributions and precursor sources (20, 32, 33). The recent NASA Atmospheric Tomography (ATom) mission provides global-scale and seasonally resolved in situ measurements of O3 and CO and a comprehensive suite of trace gases and aerosol parameters, including tracers of BB and urban emissions (34). ATom sampled the remote troposphere from the Arctic to the Antarctic over the Pacific and Atlantic Oceans using repeated vertical profiles from ∼0.2 to ∼13 km in altitude during four seasonal deployments between 2016 and 2018 (Fig. 1). Recently, the ubiquitous presence of dilute BB smoke in the remote troposphere and its significant contribution to aerosol mass loading was established using ATom observations (35). Here, we use ATom measurements to quantify the individual contributions of urban and BB emissions to O3 in the remote global troposphere using tracers specific to each source. We compare this analysis with simulations from three global CTMs that alternatively set BB and urban emissions to zero to evaluate their impact on modeled tropospheric O3.

Fig. 1.

Map of ATom flight tracks from the four seasonal global circuits colored by tropospheric O3 mixing ratios. Note that the color scale terminates at 70 ppbv of O3, and higher values are shown in red. Measurements with a strong stratospheric influence were parsed out as indicated in Materials and Methods.

Results and Discussion

Here, we analyze in situ measurements of O3, ozone precursors, chemical tracers of continental emissions, CO, and water vapor (H2O) to evaluate the influence of urban and BB emissions on O3 levels in the remote troposphere. We use the industrial solvents tetrachloroethylene (C2Cl4) and dichloromethane (CH2Cl2) as tracers of urban and industrial activity, as more than 90% of their global emissions are anthropogenic (36–38). Conversely, we use hydrogen cyanide (HCN) and acetonitrile (CH3CN), compounds emitted almost exclusively (80 to 100%) from BB, as tracers of BB emissions (39–41). All four tracers have an atmospheric lifetime ranging roughly from 2 to 6 mo (40, 42, 43), that is, long enough to identify the influence of emissions on tropospheric O3, which has a global average lifetime of several weeks (28). CO is an atmospheric pollutant released by combustion processes extensively used as a tracer of continental pollution (44) with a global mean atmospheric lifetime of roughly 1 to 3 mo (45, 46). Air masses sampled during ATom were usually within 5 to 10 d downwind of continental pollution sources (47), such that the selected tracers were subject on average to modest (<20%) loss. In this analysis, we combine HCN and C2Cl4 as our primary pair of BB and urban tracers and CH3CN and CH2Cl2 as our secondary pair based on data availability and analytical uncertainty (see details in ).

In a previous study, we used the relationship between O3 and CO during ATom to identify several distinct air mass types: well mixed and aged, stratospheric, marine, and fresh continental outflow (48). Here, we remove data with dominant stratospheric and marine influence as well as the flight segments above the conterminous United States (see Materials and Methods). We thus retain 61% of observations for analysis of the remote troposphere and report the distribution of tropospheric O3 corresponding to each quartile of C2Cl4, CH2Cl2, HCN, and CH3CN for various regions in Regionally Enhanced O. We separately investigate high O3 season (spring and summer) and low O3 season (fall and winter) for both hemispheres. Note that the boreal spring and summer corresponds to the austral fall and winter and vice versa. This seasonal cycle was captured by ATom observations (48) and is typical of tropospheric O3 globally (49). In Sources of O, we attribute O3 measured in the remote troposphere to urban and BB sources, and in Comparison of Observed and Modeled BB Influence on Tropospheric O, we compare our results with global CTMs. All acronyms are listed in .

Regionally Enhanced O3 Levels in Polluted Air Masses.

Northern Hemisphere extratropics (>20°N).

The highest tropospheric O3 levels were measured in the Northern Hemisphere (Figs. 1 and 2 ) during ATom, with median O3 mixing ratios of 59 ppbv (parts per billion by volume) in spring and summer and 44 ppbv in fall and winter (median calculated for all altitudes). The Northern Hemisphere also had the largest abundance of CO () and urban tracers (). There was little seasonal difference in the CO abundance, but BB tracer mixing ratios were greater in spring and summer, and urban tracer mixing ratios were greater in fall and winter ().

Fig. 2.

Distributions of tropospheric O3 mixing ratios (ppbv, y-axis) for each quartile of pollution tracer (x-axis) measured during ATom in four regions of the globe. The colors in the legend indicate the source of the tracers (red for BB, gray for urban). The Northern Hemisphere (NH) extratropics (panels A and B) corresponds to latitudes >20°N, the tropics (panels C and D) are defined within 20°S to 20°N, and the Southern Hemisphere (SH) extratropics (panels E and F) correspond to latitudes >20°S. The NH and SH data are further parsed into a photochemically active period (spring and summer; S&S) and a darker period (fall and winter; F&W). The box and whisker plots show the 10th, 25th, 50th, 75th, and 90th percentiles of O3 distributions. The mustard-shaded area represents regional background O3 values defined as the average O3 mixing ratio in well-mixed and aged air masses. The width of the mustard-shaded area represents the range of background O3 values obtained when using different pairs of tracers to define the well-mixed and aged air masses. The black dashed lines show regional median O3 values.

Greater O3 levels in spring and summer compared with fall and winter reflect photochemical production of O3 in both urban- and BB-influenced continental plumes lofted and transported to the remote atmosphere (44, 50–52). While O3 production is diminished in fall and winter due to reduced sunlight exposure and resulting photochemistry (), O3 is still enhanced in polluted air masses relative to the background during those seasons (Fig. 2). Parrish et al. proposed wintertime chemical destruction of O3 to explain negative O3 to CO slopes in North American polluted air masses transported over the Atlantic Ocean (53). Here, we find that O3 does not continuously increase with increasing CO in fall and winter (), but it does continuously increase with increasing urban and BB tracers (Fig. 2). Furthermore, we find that CO continuously increases with increasing levels of urban tracers but not with BB tracers (). This supports previous findings suggesting that the O3 to CO correlation can be driven by the mixing of air masses with different origins rather than O3 production in the remote troposphere (54, 55). The mixing of urban- and BB-influenced air masses affect O3 and CO levels differently, and the resulting O3 to CO correlation depends on the composition and age of the mixture.

The seasonal and latitudinal differences in background O3 reveal the contributions of photochemical O3 production outside of pollution plumes to global tropospheric ozone. Background O3, defined as the average O3 in well-mixed and aged air (air in which both urban and BB tracers are below their regional median), is 15 ppbv greater in spring and summer compared with fall and winter in the Northern Hemisphere (Fig. 2 ), a pattern not observed in the Southern Hemisphere (Fig. 2 ). Additionally, background O3 in the Northern Hemisphere spring and summer is 25 ppbv greater than in the Southern Hemisphere spring and summer but only 5 ppbv greater in fall and winter. Attributing this increased and seasonally dependent background O3 to sources is not possible with our approach.

Tropics (20°S to 20°N).

The tropical Atlantic exhibits the second highest O3 levels after the Northern Hemisphere extratropics with a median O3 of 42 ppbv. In contrast, the tropical Pacific has the lowest levels of O3 with a median O3 of 25 ppbv despite having a higher abundance of CO () and urban and BB tracers than the Southern Hemisphere extratropics (). O3 increased with pollution tracers in both tropical regions (Fig. 2 ).

In the tropical Pacific, O3 (Fig. 2) and CO () concomitantly increase with increasing urban and BB tracers, suggesting a comparable influence of these emission sources on O3 in that region (56–58). Low O3 in the tropical Pacific has been attributed to loss in the marine boundary layer (MBL), where conditions favor the destruction of O3 via photolysis and subsequent formation of OH at high humidity without concurrent production through the HOx-NOx cycle because of the low NOx levels in this region (59–62).

In the tropical Atlantic, the levels of BB tracers are second only to the Northern Hemisphere in spring and summer (). Additionally, O3 and CO both increase with increasing levels of BB tracers, a pattern not reproduced with the urban tracers (Fig. 2 and ), illustrating the overwhelming influence of African and South American BB emissions on tropospheric O3 in this region (63–65).

Southern Hemisphere extratropics (>20°S).

The Southern Hemisphere is the most pristine region of those examined here and exhibits the lowest levels of CO () and pollution tracers (). Median O3 mixing ratios were 39 and 34 ppbv in spring and summer and fall and winter, respectively. While little O3 enhancement above background was observed with increasing pollution tracers in fall and winter (Fig. 2), more active photochemistry in spring and summer () led to significantly enhanced O3 above background in polluted air masses (Fig. 2). A more pronounced increase of O3 (Fig. 2 ) and CO () with BB tracers than with urban tracers reflect the dominance of BB emissions over urban as a source of pollution in this region inhabited by only ∼5% of the world population but with equivalent BB emissions to that in the Northern Hemisphere extratropics (66, 67).

Sources of O3 to the Remote Troposphere.

To separately quantify the respective influence of urban and BB emissions on tropospheric O3, we classify all sampled air masses into four categories based on observed concentrations of tracers. For each region, polluted air parcels are either defined as urban air (urban tracer > regional median, BB tracer < regional median), BB air (BB tracer > regional median, urban tracer < regional median), or mixed pollution air (both urban and BB tracers > regional median). The remaining data (both urban and BB tracers < regional median) are considered to be well-mixed and aged air. We define the background levels of a given species (O3, CO, and the urban and BB tracers) as the average mixing ratio of that species in well-mixed and aged air. This background is regionally and seasonally dependent. Note that this definition of background differs from that used for surface ozone in regulatory applications. For example, air masses with a strong stratospheric influence have been removed in this analysis, but they are usually included in determinations of background ozone used for regulatory purposes. Each category contains roughly equal numbers of observations. The O3 distribution for each air parcel category is shown in Fig. 3. Excess O3 (denoted as positive ΔO3) attributed to each pollution source is the difference between the average O3 in polluted air (i.e., urban, BB, or mixed pollution) and that in well-mixed and aged air (i.e., regional background O3).

Fig. 3.

Distributions of tropospheric O3 (y axis) within each air mass classification (x-axis): well-mixed and aged (WMA) air, urban (UR), BB, or mixed pollution (MP) as defined in Sources of O3 to the Remote Troposphere. The Northern Hemisphere (NH) extratropics (panels A and B), the tropics (panels C and D), and the Southern Hemisphere (SH) extratropics (panels E and F) were defined as in Fig. 2. ATom observations classified using HCN and C2Cl4 or CH3CN and CH2Cl2 pairs of tracers are shown in yellow and green, respectively. Modeling results are shown by the hashed box and whisker plots with GEOS-Chem1 in violet (2 × 2.5° horizontal resolution), GEOS-Chem2 in pink (4 × 5° horizontal resolution), GFDL-AM3 in light blue, and CAM-chem in gray. The boxes and whiskers show the 10th, 25th, 50th, 75th, and 90th percentiles of O3 distributions. ΔO3 corresponds to the difference between average O3 in polluted air (i.e., UR, BB, or MP) and that in WMA air. ΔO3 is plotted with square markers for ATom observations and round markers for modeling results. The error bars correspond to the SEs of the average O3 in polluted air and in WMA air added in quadrature.

We compare ATom observations with three global CTMs: GEOS-Chem, CAM-chem, and GFDL-AM3 (). Ozone and CO are simulated along the ATom flight tracks in a base case scenario with both anthropogenic and BB emissions (simulation A), zero anthropogenic emissions (simulation B), and zero BB emissions (simulation C). Since CO is directly emitted by both fossil fuel combustion and BB, anthropogenic CO (COUR) is determined as the difference between simulations A and B, and BB CO (COBB) is determined as the difference between simulations A and C. Using COUR as the model urban tracer and COBB as the model BB tracer, we derived the modeled O3 distribution associated with urban, BB, mixed-pollution, and well-mixed and aged air and the corresponding ΔO3 following the same methodology as for the observations.

ATom observations show that O3 levels were generally more enhanced in mixed pollution air, followed by BB air and then urban air, in all regions (Fig. 3). This result is generally consistent for either set of urban and BB tracers used. Larger O3 levels in mixed pollution air usually reflect increased O3 production (68–70) from greater NOx and peroxy acyl nitrates (PANs)—a reservoir species of NOx—in mixed-pollution air compared with the BB- or urban-dominated air alone (68–71). We report in Fig. 4 regional median mixing ratios of the four most abundant reactive nitrogen (NOy) species measured during ATom: NOx, PANs, and nitric acid (HNO3) as well as particulate nitrate (pNO3). NOy concentrations were almost always greatest in mixed-pollution air followed by BB air and then urban air, similar to those for O3. The only exception was in the tropical Atlantic, where median NOy was highest in BB air, resulting in the highest ΔO3 (Fig. 2). Higher NOy concentrations in BB air than in urban air is due to generally higher NOx and PANs levels (Fig. 4), a pattern analogous to what is observed for O3 (Fig. 3). Although global inventories estimate five times more NOx emissions from fossil fuel combustion than from BB sources, only a limited fraction of the total NOx and PANs emitted in the boundary layer reaches the free troposphere (72, 73). BB emissions, on the other hand, are often convectively injected to higher altitudes, resulting in a proportionally larger amount of BB NOx than urban NOx injected into the free troposphere (74, 75). In the remote troposphere, O3 production is driven by the abundance of NOx or its reservoir species such as PANs (27, 76, 77). Injection at higher altitude and colder temperatures also increase the lifetime of thermally labile NOx reservoirs such as PANs, whose slow thermal decomposition enables sustained ozone production during long-range transport (78).

Fig. 4.

Median mixing ratios of the four main NOy species measured during ATom in four air mass types: 1) well mixed and aged, 2) urban, 3) BB, and 4) mixed pollution. NOy species are indicated with different colors as shown in the legend.

Often, the global CTMs used here predict lower O3 in polluted air masses than in well-mixed and aged air, particularly in the Northern Hemisphere and in the tropical Atlantic (negative ΔO3; Fig. 3 ). Greater levels of O3 in well-mixed and aged air imply that tropospheric O3 sources in these CTMs are overweighted toward either chemical O3 production from nonurban and non-BB emissions or stratospheric mixing. Additionally, CTMs often overestimate ΔO3 in urban air masses in the Southern Hemisphere in fall and winter and do not reproduce ΔO3 in BB and mixed-pollution air masses in the Southern Hemisphere in spring and summer (Fig. 3 ). CTMs sometimes reproduce ΔO3 in polluted air masses in the tropical Pacific, where overall O3 mixing ratios are lowest (Fig. 3), but they predict significantly [ANOVA test combined with a Tukey post hoc test with a 95% CI (79)] greater levels of O3 in well-mixed and aged air compared with observations there and in all other regions except for the Southern Hemisphere in fall and winter (Fig. 3). Discrepancies between models and observations highlight gaps in our understanding of tropospheric O3 sources.

Comparison of Observed and Modeled BB Influence on Tropospheric O3.

To approximately quantify the contributions of BB and urban emissions to O3, we first calculate their respective influence on every tropospheric air parcel sampled during ATom in which either the urban or BB tracer was above the regional background value (45 to 76% of the data; ). The fraction of influence is calculated as the ratio of the normalized excess mixing ratio of an urban and BB pollution tracer to CO using ATom observations to its emission ratio from source inventories (see Materials and Methods). A simple mixing model shows that variations in the fraction of influence of a source-specific tracer quantitatively represents its mixing with different air mass types (). Next, we calculate the amount of O3 attributed to urban and BB emissions as the product of the fraction of influence and the amount of O3 above background. Underlying assumptions and uncertainties are further discussed in Materials and Methods and in . We report the concentration-weighted average O3 attributed to urban and BB emissions per region in Fig. 5. The concentration-weighted average corresponds to the average regional O3 attributed to urban/BB emissions weighted by the urban/BB tracer mixing ratio. Additionally, we estimate the contributions to O3 from BB and urban emissions using the same simulations from the three global CTMs (Fig. 5). The difference between simulation A (base case) and B (no anthropogenic emissions) is interpreted as urban O3, and the difference between simulation A (base case) and C (no BB emissions) is interpreted as BB O3.

Fig. 5.

Concentration-weighted average O3 attributed to urban and BB emissions. ATom observations are shown in green (the two markers per symbol are from the two pairs of tracers used to identify air mass influences), and modeling results are shown in brown. The uncertainties are in (observations) and (modeling results).

Observations demonstrate consistently similar or greater O3 contribution from BB compared with urban emissions in most regions and seasons, with most confidence in the Southern Hemisphere spring and summer and tropical Atlantic. In the Southern Hemisphere and the tropical Atlantic, BB emissions contribute ∼2 to 10 times more O3 than urban emissions. In the Northern Hemisphere and the tropical Pacific, BB and urban emissions contribute similar levels of O3. The highest concentration-weighted average O3 attributed to BB emissions is found in the Southern Hemisphere spring and summer (23.0 to 24.7 ppbv) followed by the tropical Atlantic (10.8 to 12.5 ppbv). The highest concentration-weighted average O3 attributed to urban emissions is in the tropical Pacific (9.0 to 10.0 ppbv) followed by the Northern Hemisphere spring and summer (3.3 to 7.5 ppbv). The O3 attribution has a large uncertainty (), and the methodology applied to derive these numbers is sensitive to many parameters that also have large uncertainties (e.g., background value, emission ratio; ). However, we apply similar assumptions and parameters to calculate O3 from urban and BB emissions, which should reduce the uncertainty in the ratio of O3 attributions to these sources. Consistent results for two sets of urban and BB tracers further support this methodology (Figs. 3 and 5). Additionally, an alternate method consisting of removing air parcels with significant BB or urban influence to evaluate the shift in the regional O3 distribution also points to a similar or stronger influence from BB than urban emissions on tropospheric O3 (). Indeed, removing air masses with large BB or mixed pollution influence generally reduces O3 by a few ppbv, whereas removing air masses with large urban influence generally shifts the O3 distribution toward higher values.

None of the models adequately represents the strong influence of BB emissions on tropospheric O3. Only one model, GEOS-Chem1, attributed similar O3 enhancements to BB and urban emissions but only in the Southern Hemisphere and the tropical Atlantic. The other two models, as well as GEOS-Chem1 in the Northern Hemisphere and the tropical Pacific, attribute a factor of 2 to 10 more O3 to urban than BB emissions. In general, all models ascribe similar or more O3 to urban emissions than that derived from observations. Since these observation-derived O3 enhancements are relative to background O3, which itself is increased by pollution outflow (80), they are lower estimates of the O3 contribution from BB and urban emissions. Accounting for contributions from urban and BB emissions to observed background O3 could reduce the discrepancy with modeled urban O3 but would likely further increase the gap with modeled BB O3.

Discrepancies in the attribution of tropospheric O3 to BB emissions between observations and model results may arise from a misrepresentation of 1) BB emissions, 2) smoke chemistry, and 3) plume injection height in global models. A correct representation of the injection altitude of smoke in the troposphere is crucial to better predict the impact of BB emissions on the composition of the atmosphere (81). Zhu et al. recently proposed a new global BB emissions injection scheme (not used here by GEOS-Chem), which resulted in significantly better agreement of the simulated vertical profiles of PAN and CO with observations over the North American boreal regions (82). Uncertainty in BB emission inventories has been suggested as a potentially large source of error in global modeling of tropospheric O3 by the recent Tropospheric Ozone Assessment Report modeling analysis (21). Furthermore, BB emission inventories rely on satellite images of burned areas with coarse spatial resolution and thus significantly underestimate the contribution of undetected small fires (83). Finally, O3 chemistry in BB plumes is complex and is a topic of intensive current research. Recent findings on the emissions and chemical evolution of NOy species (71, 84, 85), VOC emissions (18), and O3 formation in smoke plumes (86, 87) may improve constraints for global models and enable a better representation of chemical mechanisms in smoke.

Conclusion and Implications.

Global-scale in situ observations of O3, CO, and pollution tracers in the remote troposphere demonstrate that both urban and BB emissions strongly influence global tropospheric O3. Air parcels with pollution from both sources exhibit the greatest O3 enhancements followed by air influenced by BB and then urban emissions in all regions of the globe. A similar pattern is observed for NOy species, which suggests significant O3 production due to continental pollution in the remote troposphere. All global CTMs used here overpredict O3 in well-mixed and aged air masses, usually leading to an underprediction of the excess O3 in polluted air masses when compared with the observations. Observations are analyzed to show that in the Northern Hemisphere, where most urban emissions occur, and in the tropical Pacific, O3 attributed to BB was within a factor of two of that attributed to urban sources, while in the Southern Hemisphere and in the tropical Atlantic, BB accounted for 2 to 10 times the ozone from urban sources. Conversely, all global CTMs used in this study ascribe a factor of ∼1 to 10 more O3 from urban than from BB emissions for the same air parcels, possibly arising from uncertainties in BB emission inventories, chemical mechanisms for O3 production, plume injection heights, or all three.

On the global scale, anthropogenic activity has increased the tropospheric ozone burden by ∼40% since 1850 (28, 88), with modeled increases at northern midlatitudes of ∼70% since 1850 and 40% since 1950 (89), in broad agreement with limited observations from the early and mid-20th century (90). In contrast, inventories suggest that global BB emissions have not changed substantially since preindustrialization, as the observed decrease in African emissions since about 1950 has been partially compensated for by an increase in emissions from South America and Asia (66). Nevertheless, there is mounting evidence that BB emissions will play an increasing role in the future composition of the atmosphere due to longer fire seasons (91), more intense fire activity (92), and emissions reductions from fossil fuel combustion. Characterizing these emissions and their impacts at all scales is a subject of intense current public and scientific interest (93–97). Accurate representation of BB emissions and related atmospheric chemistry in global CTMs will be crucial to predicting future levels of tropospheric O3 and its role in the radiative forcing of climate and air quality. The increased influence of BB O3 on regional backgrounds compromises the mitigation of O3 through local emissions reductions to meet air quality standards (98). Estimates of O3 radiative forcing are dependent on accurate understanding of relevant emission sources and chemistry and may depend more strongly on BB emissions than currently predicted (99). The observational analysis of BB influence at the global scale suggests an incomplete representation of global ozone in current models and points toward several mechanisms requiring further investigation to address these uncertainties.

Materials and Methods

Data Parsing.

We used the 2-min merged ATom dataset in this study (34). We removed flight segments over the continental United States to limit the influence of regional emissions. Next, we filtered out marine and stratospheric influences by retaining measurements with O3/H2O between 0.003 and 1 ppbv ⋅ ppmv−1 (parts per million by volume) . This resulted in 61% of the ATom dataset considered as remote tropospheric air. The thresholds of 0.003 and 1 ppbv ⋅ ppmv−1 were chosen based on the inflections in the O3 to O3/H2O correlation (). A similar filter based on the modeled O3 and H2O was applied to the results from the CTMs. The sensitivity to these filters is further discussed in and illustrated in . More details about the instrumentation during ATom and the modeling simulations used in this work can be found in .

Fraction of Influence.

We quantified the respective influence of BB and urban emissions on each air parcel by first calculating the normalized excess mixing ratio (NEMR) of BB and urban tracers according to Eq. 1:where NEMRX is the normalized excess mixing ratio of compound X (i.e., X = HCN, CH3CN, C2Cl4, or CH2Cl2) to CO, and ΔX (ΔCO) is the difference between the mixing ratio of compound X (CO) and its background level. Background levels were defined as the average mixing ratio in well-mixed and aged air masses, defined as air where both urban and BB tracers are lower than their regional median values (see Sources of O3 to the Remote Troposphere).

The respective influence of urban and BB emissions (FX) was then calculated as the ratio of the NEMR of compound X to the average emission ratio of compound X (ERX) as follows:

The determination of ERs for each pollution tracer is further discussed in and illustrated in . Urban and BB fractions of influence were calculated for air masses with urban or BB tracer levels above their background values, thus limiting the analysis to air parcels with quantifiable continental pollution. The proportion of such air masses is given for each region in . We present in a histogram of the fractions of influence in the tropical Atlantic as an example. The fractions of influence are distributed around values that reflect the most commonly observed influence from each emission type. In the tropical Atlantic, we show a larger influence of BB compared with urban emissions on polluted air parcels, consistent with our understanding of pollution sources in the area. Further details on the calculation of the fraction of influence are given in .

O3 Attribution Calculation.

We attribute tropospheric O3 above background to urban and BB emissions in polluted air masses as follows:

Then, we calculate the weighted-average O3 attributed to urban (BB) emissions in each region, which is the average of attributed O3 in each air parcel weighted by the urban (BB) tracer mixing ratio, as follows:where x is the weighted-average attributed O3, x is the attributed O3 in the air parcel i, and w is the weight associated with each value of O3 attribution equal to the tracer mixing ratio.

One of the assumptions inherent to this O3 attribution is that O3 enhancements are linearly correlated with urban and BB tracers. In the remote troposphere, O3 production is linearly driven by NOx availability (77). However, a significant fraction of tropospheric O3 may reflect export from the polluted continental boundary layer (73), where O3 production is nonlinearly determined by the NOx to VOCs regime (26). The balance between export from the boundary layer and in situ production of tropospheric O3 is thus key to assess the validity of this assumption but is not constrained at the global scale. Here, we consider the simultaneous increase of O3, CO, and NOy species with urban and BB tracers (Figs. 2 and 4 and ) as evidence to support a linearity assumption. This simple calculation should thus provide a useful estimate of the average tropospheric O3 enhancement from each source. We further discuss and illustrate the uncertainty associated with this term in . A caveat to this methodology is that it assumes similar ozone production efficiency (OPE) from urban and BB emissions. The range of measured ΔO3 to ΔCO, a proxy of OPE (44), in urban and BB plumes span similar values (∼0 to 1 ppbv ⋅ ppbv−1) at the regional scale (19, 63, 100, 101). In our analysis, O3 detrained from the continental boundary layer to the remote troposphere is integrated over large spatial scales, and the event-scale differences in OPE is likely smoothed out in the process. However, differing OPEs would lead to a corresponding overestimation of the attributed O3 from the source with the lower OPE.