Impact of stratospheric air and surface emissions on tropospheric nitrous oxide during ATom

. We measured the global distribution of tropospheric N 2 O mixing ratios during the NASA airborne Atmospheric Tomography (ATom) mission. ATom measured concentrations of ~300 gas species and aerosol properties in 647 vertical profiles spanning the Pacific, Atlantic, Arctic, and much of the Southern Ocean basins, from nearly Pole to Pole, over four seasons (2016–2018). We measured N 2 O concentrations at 1 Hz using a Quantum Cascade Laser 40 Spectrometer. We introduced a new spectral retrieval method to account for the pressure and temperature sensitivity of the instrument when deployed on aircraft. This retrieval strategy improved the precision of our ATom QCLS N 2 O measurements by a factor of 3 (based on the stdev. of calibration measurements). Our measurements show that most of the variance of N 2 O mixing ratios in the troposphere is driven by the influence of N 2 O-depleted stratospheric air, especially at mid and high latitudes. We observe the downward propagation of lower N 2 O mixing ratios (compared to 45 surface stations) that tracks the influence of stratosphere-troposphere exchange through the tropospheric column down to the surface. The highest N 2 O mixing ratios occur close to the equator, extending through the boundary layer and free troposphere. We observed influences from a complex and diverse mixture of N 2 O sources, with emission source types identified using the rich suite of chemical species measured on ATom and with the geographical origin calculated using an atmospheric transport model. Although ATom flights were mostly over the oceans, the most prominent N 2 O 50 enhancements were associated with anthropogenic emissions, including industry, oil and gas, urban and biomass burning, especially in the tropical Atlantic outflow from Africa. Enhanced N 2 O mixing ratios are mostly associated with pollution-related tracers arriving from the coastal area of Nigeria. Peaks of N 2 O are often associated with indicators of photochemical processing, suggesting possible unexpected source processes. In most of the cases, the results show the difficulty of separating the mixture of different sources in the atmosphere that may contribute to uncertainties in the 55 N 2 O global budget. The extensive data set from ATom will help improve the understanding of N 2 O emission processes and their representation in global models.


Introduction
Nitrous oxide (N2O) is a powerful greenhouse gas and, due to its oxidation to NOX, a major contributor to both stratospheric ozone loss and to passivation of stratospheric oxy-halogen radicals (Forster et al., 2007;Ravishankara et al., 2009). The increasing rate of 0.93 ppb yr -1 of atmospheric N2O since the industrial revolution implies significant (~30%) imbalance between emission rates and destruction in the stratosphere. Seasonal cycles in tropospheric N2O are driven by both stratosphere-to-troposphere exchange and by surface emissions (Nevison et al., 2011;Assonov et al., 2013;Thompson et al., 2014a). Most N2O emissions are attributed to microbial nitrification and denitrification in natural and cultivated soils, freshwaters and oceans, plus emissions related to human activities such as biomass burning and industrial emissions (Butterbach-Bahl et al., 2013;Saikawa et al., 2014;Thompson et al., 2014a;Upstill-Goddard et al., 2017;WMO, 2018).

180
Much effort has been made to reduce the uncertainties in the individual components of the N2O global budget (e.g., Tian et al., 2012Tian et al., , 2020Xiang et al., 2013;Thompson et al., 2014a, b;Ganesan et al., 2020;Yang et al., 2020). Recent estimates of global total N2O emission to the atmosphere from bottom-up and top-down methods average 17 Tg N yr -1 (12.2-23.5 from bottom-up analysis, and 15.9-17.7 Tg N yr -1 from top-down approaches, Tian et al., 2020). The most recent estimates of the global ocean emission of N2O range between 2.5 and 4.3 Tg N yr -1 ( ~ 20% of total emissions), and temporal heterogeneity (Nevison et al., 1995(Nevison et al., , 2005Ganesan et al., 2020;Yang et al., 2020). According to Tian et al. (2020), anthropogenic sources account for ~ 43% of the global N2O emissions (7.3 Tg N yr -1 ), with industry and biomass burning emissions estimated to be 1.6 to 1.9 Tg N yr -1 respectively (Syakila and Kroeze, 2011;Tian et al., 2020) and the rest from agriculture. N2O emissions from biogenic sources and fires in Africa are estimated at 3.3 ± 1.3 Tg N2O yr -1 (Valentini et al., 2014). Agricultural N2O emission estimates (up to ~ 37%) range between 2.5 and 5.8 Tg 210 N yr -1 , and between 4.9 and 6.5 Tg N yr -1 in the case of natural soils (Kort et al., 2008;Syakila and Kroeze, 2011;Tian et al., 2020). Recent estimates of N2O emissions from fertilized tropical and subtropical agricultural systems are 3 ± 5 kg N ha −1 y −1 (Albanito et al., 2017). Most of these estimates are derived from short-term local-scale in-situ measurements and are difficult to extrapolate with confidence to large regions or to the globe.
In the atmosphere, N2O is destroyed by oxidation (10%, O( 1 D) reaction) and photolysis (90%, 190-230 nm 215 photolysis) in the upper stratosphere (> 20 km altitude; SPARC, 2013), which makes it a good candidate for tracing the air exchange between the stratosphere and the troposphere (Hintsa et al., 1998;Nevison et al., 2011;Assonov et al., 2013;Krause et al., 2018). Atmospheric models tend to underestimate the inter-hemispheric N2O gradient, which Thompson et al. (2014a) attribute to an overestimation of N2O emissions in the Southern Ocean, an underestimate of the Northern Hemisphere emissions, and/or an overestimate of stratosphere-to-troposphere exchange in the Northern 220 Hemisphere. Overall, the largest uncertainties of modelled N2O emissions are found in tropical South America and South Asia (Thompson et al., 2014b).
We present atmospheric N2O altitude profiles at high temporal resolution collected during the NASA Atmospheric In this work we focus on the measurements taken during January-February 2017 (ATom-2), September-October 2017 (ATom-3), and April -May 2018 (ATom-4) (no Quantum Cascade Laser spectrometer (QCLS) N2O data are available for ATom-1 in Aug. 2016). The motivation for this paper is twofold. Firstly, we present a new retrieval strategy to account for the pressure and temperature dependence of laser-based instruments, specifically for the use of 235 quantum cascade laser spectrometers on aircraft. Secondly, we report on the global distribution of N2O from the surface to 13 km, and examine the processes contributing to the variability of tropospheric N2O based on the vertical profiles of N2O and a broad variety of co-variate chemical species and aerosol properties.
2 Instrument specifications, spectral analysis and calibration
QCLS provides continuous (1 Hz) measurements of N2O, methane (CH4) and carbon monoxide (CO) using two thermoelectrically-cooled pulsed quantum cascade lasers, a 76 m pathlength multiple pass absorption cell (~ 0.5 L volume), and two liquid-nitrogen cooled solid-state HgCdTe detectors. All these components are mounted on a temperature-stabilized, vibrationally isolated optical bench. The temperature in QCLS is controlled by Peltier elements 260 coupled with a closed-circuit recirculating fluid kept at 288.0 ± 0.1 K. QCLS measures CH4 and N2O by scanning the spectral interval of 1275.45 ± 0.15 cm -1 . A second laser is used to scan CO at 2169.15 ± 0.15 cm -1 . The supply currents to QCLS are ramped at a rate of 3.8 kHz to scan the laser frequency for 200 channels (steps in frequency) in laser 1 and 50 channels in laser 2, with an extra 10 channels to measure the laser shut off (zero-light level). The spectra and fit residual for CH4, N2O and CO are shown in Fig. S1 of the Supplement. Mixing ratios are derived at a rate of 1 Hz by a 265 least-squares spectral fit assuming a Voigt line profile at the pressure and temperature measured inside the sample cell and using molecular line parameters from the HIgh-resolution TRANsmission molecular absorption database (HITRAN, Rothman et al., 2005). The temperature and pressure inside the cell are monitored with a 30 kΩ thermistor and a capacitance manometer (133 hPa full scale), respectively.
During sampling, the air passes through a 50-tube Nafion drier to remove the bulk water vapor. A Teflon diaphragm 270 pump downstream of the cell reduces the air pressure to ~60 hPa. Both ambient air and calibration gases pass through a Teflon dry-ice trap to reduce the dew point to -70 °C. After ATom-1, we added a bypass between the inlet and the instrument to increase the flushing rate of the inlet and inlet tubing. The calibration sequence includes 2 minutes of Ultra-High Purity zero air, followed by 1 minute each of low-and high-mixing ratio gases every 30 minutes (see Fig.   S2). During ATom-1 and -2, we measured zero air every 15 minutes, and every 30 minutes during ATom-3 and -4. A data logger (CR10X, Campbell Scientific) was used to automate the sampling sequence. The CR10X controlled the pressure controller on the cell and managed the data transfer.
We use gas cylinders traceable to the National Oceanic and Atmospheric Administration World Meteorological

285
The high-mixing ratio gas cylinder contained 399.1 ± 0.3 ppb of N2O, 2182.5 ± 0.3 ppb of CH4 and 192.8 ± 0.5 ppb of CO, respectively. Detailed information on calibrations of the gas cylinders used during ATom are in Table S1 of the   Supplement. QCLS also measures carbon dioxide (CO2) in a separate unit. Detailed information about QCLS CO2 measurements can be found in Santoni et al. (2014).

Spectral analysis and calibration
The QCLS was damaged during shipping to the deployment site before the start of ATom-1, and the resulting alteration in the optical alignment modified the sensitivity of the instrument to temperature and pressure changes during aircraft  Fig. S3 for ATom), but drifts were observed during altitude changes due to the effects of changes in cabin pressure and temperature on the spectral location of interference fringes that arise in the optical path outside the sample cell. In addition, flight altitude changes could mechanically stress the optical elements surrounding the cell, further modulating fringes or changing the shape of the detected laser intensity profile. These spectral artifacts 300 ultimately reduced the accuracy of mixing ratios retrieved from spectral fitting. The spectral artifacts most strongly affected the measurements of CH4 and N2O. Several post-processing methods using the TDL-Wintel software were explored to improve the precision and accuracy of ATom QCLS N2O data, most with little success. Since the measured spectra were all saved, it is possible to re-fit the data with different fit parameters. A limited number of interference fringes may be included in the set of fitting functions. However, none of the previously used full re-fitting strategies 305 significantly improved the data accuracy.
We have achieved significant improvement in the precision and accuracy of the ATom QCLS N2O data using a new method dubbed the "Neptune algorithm", developed by Aerodyne Research, Inc., and that has been further developed and applied to the data sets described here. Using this algorithm, the precision of the retrieved N2O data measured with the damaged QCLS, was similar to that reported in HIPPO. The Neptune algorithm generates corrections 310 to the mixing ratios retrieved from the original fits by associating specific spectral features with anomalies in retrieved mixing ratios observed during calibrations, i.e., during intervals when the mixing ratios are held constant. The spectral baseline is defined as the spectral channels outside the boundaries of the spectral lines of the target gas. Fluctuations in the spectral baselines are quantified for the entire data set by means of principal component analysis (PCA). PCA provides an efficient description of the spectral fluctuations, naturally producing an ordered set from strongest to 315 weakest of orthogonal vectors (spectral forms), each with an amplitude history spanning the data set. The PCAs are defined by an optimization procedure during calibrations, when mixing ratio fluctuations are designed to be ~0. The finite fluctuations in retrieved mixing ratios during calibrations are fit in the spectral space of the baseline as linear Deleted: temperature and combinations of the leading PCA vector amplitudes, creating a linear combination of amplitudes of spectral fluctuations that predict errors in the mixing ratios for each gas for an entire flight. The error-producing linear combination of amplitudes of PCA spectral fluctuations produces a full set of anomaly estimates, which is subtracted from the retrieved mixing ratios during the flight. The computational time for a 10-hour long dataset is only seconds, so variations in the algorithm's parameters (i.e. how many PCAs are retained) can be optimized rapidly.
The Neptune-PCA analysis improved the overall precision by a factor of 4 for CH4 and a factor of 3 in the case of 325 N2O with respect to the precision of the original retrievals, as measured by the standard deviation of retrieved mixing ratios during calibrations. The repeatability of the retrieved calibrations was 0.2 ppb for N2O and 1 ppb for CH4 (Fig.   S4). The laser path of the CH4/N2O laser was realigned between ATom-1 and -2 and the Neptune retrieval was applied to CH4 and N2O measurements corresponding to the ATom-2, -3 and -4 deployments. Mixing ratios of CH4 and N2O could not be retrieved during ATom-1 because light levels were too low for the CH4/N2O laser due to the damage-330 induced misalignment.
The steps involved in the Neptune correction process were as follows.
1) We paired the mixing ratio records with the corresponding spectra (1-s resolution) for each species (CH4 and N2O).
2) We grouped the mixing ratios and spectra by type as calibrations (zeros, low span and high span) and air samples, and in time. The spectral data were thus arranged in an array, with point number in the spectrum as x, and spectrum number as y. We calculated an average spectrum for each group type and subtracted these from each individual spectrum within a group.
3) We zeroed-out the spectral arrays at the positions of the absorption lines to concentrate on the fluctuations observed in the baseline and to prevent the PCA from finding line-depth fluctuations as relevant vectors during the 340 calibrations. Some degree of smoothing (in x) was applied to the subtracted spectra, so that high-frequency fluctuations, which have little influence on the mixing ratio determination, are not represented. An example of such a processed spectral array is shown in Fig. 1a.

4)
We applied PCA to the whole line-zeroed spectral array to evaluate the fluctuations. PCA is applied in two steps: multiply the spectral array by its transpose, to generate an autocovariance array; and then perform singular value 345 decomposition on the autocovariance array. The PCA generated an efficient description of how the baseline of the spectrum changed with cabin pressure and temperature. The description of spectral fluctuations is made in terms of a set of products of vectors and amplitudes.

5)
We fit the spectra to the PCAs to express mixing ratio fluctuations during the set of calibrations and zeros as a linear combination of PCA vector histories. The number of vector histories that we included in the fit typically is limited 350 to less than 30, because the weaker PCA amplitudes tend to just describe random noise.
The linear combination of amplitudes that links spectral fluctuations in the baseline to mixing ratio fluctuations during calibrations, was then applied to the full data set. That generated the retrieval errors for uncalibrated mixing ratios, for the whole time series. We subtracted the errors from the initial retrievals from TDLWintel-QCLS software and computed calibrated mixing ratios using the corrected retrievals for both calibrations and samples. An example of

365
We evaluated N2O mixing ratios measured by QCLS against three other instruments measuring N2O on the NASA DC-8 aircraft during ATom. In addition, we compared the set of 4 airborne measurements to data from the flask sampling network at ground stations from the NOAA Global Monitoring Laboratory (GML, https://www.esrl.noaa.gov/gmd/) to evaluate the differences between the airborne data and the ground-based measurements in the NOAA reference network.

375
We compared QCLS, PANTHER and UCATS in 10s intervals as provided in the ATom merged file,

Comparison between airborne and surface measurements of N2O
We evaluate the traceability of lower-troposphere N2O mixing ratios by ATom by comparing the four airborne instruments with the surface measurements of N2O from the NOAA flask sampling network. If during a flight, a surface station was encountered within a latitude range of 5 degrees north and south with respect the flight track, that surface station was used in the study.
A mean value of N2O within that latitude grid of +/-5 degrees and between 1 to 4 km altitude of instrument was compared with the mean N2O at the surface station observed between +/-5 days relative to the flight (due to the nondaily frequency of flask samples). We chose the altitude range between 1 to 4 km to agree with the low free troposphere conditions that characterized most of the selected ground stations. Information about the surface stations used here is shown in Table

420
The vertical profiles of N2O from ATom provide a global overview of the N2O distribution in the troposphere with observations over the Pacific and Atlantic basins. For this study we do not include data collected over and close to land.
Deleted: at common sampling locations. Information about the surface stations used here is shown in In ATom, N2O ranged between 280 and 335 ppb over the oceans. In each season, the lowest N2O mixing ratios are observed at high latitudes (HL, > 60°) in the UT/LS (8-12.5 km) in air transported downward from the stratosphere.
Tropospheric N2O tends to increase towards northern latitudes as a result of higher anthropogenic emissions in the Northern Hemisphere relative to the Southern Hemisphere. More details on the variability of N2O mixing ratios along 445 the tropospheric column are described in Section S1.
We study the impact of N2O sources and stratospheric air on the N2O column based on the anomalies (enhancements and depletions) we observed in the airborne N2O mixing ratios relative to the N2O "background" defined here as the NOAA-MBL product. We use the NOAA-MBL product to constrain a latitudinal gradient of N2O mixing ratios at the surface for each deployment. These data have been widely used to estimate the N2O background (Assonov  Fig. 3. The data describe the overall homogeneity of N2O in the troposphere (30% of the anomalies ranged between ± 0.5 ppb). We consider the ± 0.5 ppb interval to account for the day-to-day and seasonal variability of N2O. Episodes of N2O depletion (< -0.5 ppb) related to the influence of 455 stratospheric air are observed in 53.5% of the aircraft samples during ATom-2 to -4, whereas episodes of N2O enhancement (> 0.5 ppb) related to the contribution of N2O sources account for 16.5% of the calculated anomalies.
Trajectories and associated surface influence functions were computed using the Traj3D model (Bowman, 1993) and wind fields from the National Center for Environmental Prediction Global Forecast System (NCEP GFS). Model trajectories were initialized at receptors spaced one minute apart along the ATom flight tracks, and followed backwards 460 for 30 days, and reported at 3-hour resolution. From these trajectories we calculated the surface influence for each receptor point (footprints in units of ppt/(nmol m -2 s -1 )). The footprint can be convolved with a known flux inventory of a non-reactive gas to calculate the expected enhancement/depletion of that gas for each receptor point.

465
We observed the strongest depletions (> 5 ppb) in N2O mixing ratios at high latitudes and altitudes, consistent with stratospherically influenced air (Fig. 3). Stratosphere-troposphere exchange processes allow stratospheric-depleted N2O to be distributed throughout the troposphere. The NOAA surface network shows a seasonal minimum of N2O 2-4 months later than the stratospheric polar vortex break-up season. This seasonal minimum is observed at the surface around May in the southern hemisphere and around July in the northern hemisphere (see Fig. S8 Fig 3e). The N2O depletion is likely the result of stratospheric air being moved downwards by the BDC and trapped by the polar vortex, with a more pronounced effect in the southern hemisphere where the polar vortex is stronger. These results support previous work suggesting that downward transport of stratospheric air with low N2O exerts a strong influence on the variance of tropospheric N2O mixing ratios (Nevison et al., 2011;Assonov et al, 2013).
The impact of stratosphere-to-troposphere transport can be studied by combining information on tracers of 480 stratospheric air such as ozone (O3 from the NOAA -NOyO3; Bourgeois et al., 2020), sulfur hexafluoride (SF6 from PANTHER), CFC12 (from PANTHER) and carbon monoxide (CO from QCLS). These tracers are usually used either because they are strongly produced in the stratosphere (e.g. O3) or because they are tracers of anthropogenic emissions in the troposphere with a strong stratospheric sink (e.g., CO, SF6 and CFC12). In addition, meteorological parameters such as potential vorticity (PV), the product of absolute vorticity and thermodynamic stability (PV was generated by 485 GEOS5-FP for ATom) can be used to trace the stratosphere-to-troposphere transport.
Overall, the interhemispheric gradient of N2O is much smaller than that of CO and SF6 (Fig. 4), but the difference for each species is driven by larger anthropogenic emissions in the northern hemisphere. The tracer-tracer correlations shown in Fig. 4 show different patterns. The linear trend between N2O and O3 or CFC-12 highlights the role of depletion (N2O and CFC-12) and production (O3) in the stratosphere (Fig. 4a1, 4a4). When N2O is plotted against the 490 anthropogenic tracers, CO and SF6, two distinct trends are observed. Tropospheric N2O can be identified as the horizontal band containing high N2O (> 328 ppb) and variable CO and SF6, whereas the vertical band with variable N2O and small changes in CO and SF6, shows the mixing between tropospheric air and stratospheric air depleted in N2O ( Fig. 4a1-4a3). The N2O versus CO plot shows an L-shaped (bimodal) curve similar to those typically observed on O3-CO correlations during events of stratosphere-to-troposphere airmass mixing (Fig. 4a2, Krause et al., 2018). A 495 quasi-vertical line in the N2O-CO plot (e.g. constant CO) is indicative of a strong impact of stratospheric air, where CO shows the stratospheric equilibrium mixing ratio (Krause et al., 2018). The lower the CO background, the greater the influence of the stratospheric air during the airmass mixing (North Atlantic high latitudes in Fig. 4a2) and vice versa. A strong correlation is also indicative of rapid mixing between the two air masses. During ATom, the strongest impact of stratospheric air was observed in the Pacific mid and high latitudes in February (ATom-2) and in the Atlantic 500 in May (ATom-4, Fig. S11). At the Pacific northern mid and high latitudes (NMHL > 30º N), we found a consistent linear relationship between N2O and O3, with a relatively constant N2O/O3 slope (-0.05 to -0.04) during all seasons.
Linear correlations between N2O and CFC-12 highlight the dominant influence of stratospheric air depleted in these two substances in the range of mixing ratios observed at mid and high latitudes (Fig. S11).
During spring, the mid-latitudes are strongly impacted by stratospheric air due to the occurrence of tropopause 505 folds and cut-off lows to the south of the westerly subtropical jets (Hu et al., 2010 and references therein). The stronger depletion of N2O mixing ratios observed over the Atlantic relative to the Pacific during spring is due to a greater number of deep stratosphere-to-troposphere transport events at middle latitudes in the region between May and July ( Fig. 3e; Deleted: We observed a strong depletion in N2O concentrations during October 2017 at high southern 510 latitudes (SH, Fig. 3c), when the Antarctic polar vortex has broken up and the period of maximum stratosphere-totroposphere transport was ending at Northern high latitudes. Polar vortex break-up allows depleted levels of N2O to be distributed and to persist in the troposphere. During February 515 2017, 35% of the N2O observations over the Atlantic were lower than seasonal background defined from the NOAA-MBL product (i.e., N2O differences < -0.5 ppb) and 51% over the Pacific. During October 2017, depleted N2O accounted for 55% of the observations over the Atlantic and 520 46% over the Pacific, whereas during May 2018, the percentage was 53% and 49% in the Atlantic and Pacific basins, respectively. In the Arctic, the strongest signal of N2O depletion above the surface was observed in May. These results support previous work suggesting that 525 downward transport of stratospheric air with low N2O exert a strong influence on the variance of tropospheric N2O mixing ratios (Nevison et al., 2011;Assonov et al, 2013). Bands of depleted N2O begin to reach the surface in March-April in the Southern Hemisphere (ATom-4, Fig. 3b and e) arriving 530 as late as August-September in the Northern Hemisphere (ATom-3 Fig. 3c and f). Surface seasonal minima therefore occur 2-4 months later than the polar vortex break-up (see  (Fig. 4b1-4b4). The correlations between N2O and PV, and the similarities with CFC-12,

565
indicate that stratosphere-to-troposphere exchange leads to variations of tropospheric N2O up to 10 ppb in the higher latitudes for the altitudes covered during the flights. This influence is notably larger than the 2-4 ppb enhancements associated with regional emissions (see below).  Several N2O peaks were observed together with enhancements of H2O2 and PAA, which are primarily formed from chemistry occurring in the atmosphere. For the altitude range 2-4 km, regressions produced r 2 > 0.7 for 16 profiles of 590 N2O vs. H2O2 and 15 profiles of N2O vs. HCN (a tracer for combustion of biomass), but only three such profiles produced these strong associations for both H2O2 and HCN in common. Some of these profiles, shows also correlated enhancements of SO2 and NO (9 profiles r 2 > 0.6). This result raises the question as to whether globally significant production of N2O may be occurring in heterogeneous reactions involving SO2, NO redox chemistry and HONO nearby Deleted: e 595 Deleted: h Deleted: Trajectories and associated surface influence functions were computed using the Traj3D model (Bowman, 1993) and wind fields from the National Center for Environmental Prediction Global Forecast System (NCEP 600 GFS). Model trajectories were initialized at receptors spaced one minute apart along the ATom flight tracks, and followed backwards for 30 days, and reported at 3-hour resolution. From these trajectories we calculated the surface influence for each receptor point (footprints in units of ppt/(nmol m -2 s -to strong sources of reactive pollutants, as has been observed in heavily polluted atmospheres (Wang et al., 2020), or as theorized to occur in the plumes of refineries or power plants (e.g. Pires and Rossi, 1997).

620
In most cases, because we were sampling in the middle of the oceans and not over the source regions, the distinction between the different sources contributing to the observed N2O enhancements is not possible. We also observe that the impact of the different sources to N2O mixing ratios is regionally dependent. Here we describe, with some examples, the sources contributing to the major enhancements of N2O observed during ATom by oceanic regions, although we cannot pinpoint precisely the source processes.

N2O enhancements over the Pacific
Episodes of N2O enhancements were frequently observed at the southern Pacific Ocean mid latitudes, linked by the associated footprints to emissions over the continents. In this region, N2O enhancements are predominantly associated with airmasses with enhanced H2O2, PAA and CO. For example, consider Fig. 5 showing data from profile 12 on 3

630
May 2018, at 49.5-50° S near the Dateline. A distinct peak in N2O of 1 ppb amplitude, at 1700 m altitude, is significantly correlated with enhancements in CH3CN. These associations and the footprints suggest a regional contribution from fuel types from the industrial zone of Australia (Fig. 5c) which is also confirmed by the aerosol characterization from PALMS (not shown for brevity). In this profile, close to the surface, the lowest QCLS N2O mixing ratios agree with the NOAA MBL N2O (dashed line in Fig. 5b). At higher altitudes (2.5-6 km), strong correlations between N2O, H2O2, 635 PAA, CO and HCN, but not SO2, suggest the influence of biomass burning from central Australia (3-5 km) and from South America (6 km) ( Fig. 5b and 5c middle and right panels, and Fig. S11f). The relatively low mixing ratios of short-lived trace gases (PAA, H2O2 and PM1 aerosols with lifetimes ranging from hours to few days), and the surface influence based on the back trajectories (Fig. S13a), indicate that most of these profiles sampled significantly aged air masses transported for extended periods over the South Pacific.

640
At the equatorial Pacific, episodes of N2O enhancements were frequently associated with a mixture of potential marine, industrial and biomass burning emissions. Atmospheric potential oxygen (APO) is a tracer primarily of oxygen exchange with the oceans, defined as deviations in the oxygen-to-nitrogen ratio (δ(O2/N2)) corrected for changes in O2 due to terrestrial photosynthesis and respiration, and also mostly for influences from combustion (Stephens et al., 1998), Here δ(O2/N2) is the deviation in O2:N2 ratio (per meg), 1.1 is an approximation to the O2:CO2 ratio for photosynthesis and respiration, XO2 is the mole fraction of O2 in dry air, and XCO2 is the mole fraction of CO2 in the air sample (dry, µmol mol -1 ). Since APO primarily tracks oxygen exchange between the ocean and the atmosphere,

665
An example is shown in Fig. 6 for 1 May 2018. We observed high correlation between N2O and APO (r 2 = 0.66) between 0 and 4 km altitude. At these altitudes we also see enhancements in dibromomethane (CH2Br2)

N2O enhancements over the Atlantic
The Atlantic basin ATom flights saw much more continental influence than the Pacific. Around 30° N, in the North Atlantic during winter, we observe small enhancements of N2O that contrast with the overall influence of 685 stratospheric air on the tropospheric column (AT-2, Fig. 3d). The contribution is much higher during the fall season (AT-3, Fig. 3f). Several episodes of N2O enhancements are associated with enhancements of CH4, CO, and HCN. We also observe some episodes where N2O increases while CO2 decreases ( The influence of different regions on the N2O mixing ratios over the Atlantic is shown on 14 May 2018 (Fig. 7).
This profile shows the contribution to tropospheric N2O from western Europe pollution transported down over the In the Atlantic during ATom 2 (Feb 2017 ; Fig 9), the largest N2O enhancement is attributed to African agriculture (peaking at 2 ppb at 2 km), with smaller but significant influence from Asia and Europe (0.5 ppb each at 2-4 km, Fig. 725 S14). The observed and modeled N2O enhancements agree within an order of magnitude for the profile, but the model underestimates the high altitude (4-7 km) N2O enhancement by <1 ppb and overestimates the lower altitude enhancement (2-4 km) by ~1 ppb. The difference in N2O enhancement could be due to a strong latitudinal gradient in N2O across this profile or the timing of emission of N2O sampled along this single profile compared to a monthly mean estimate from the inventory. Strong correlations between N2O and HCN (r 2 = 0.95), CO and CH3CN suggest a source of N2O from burning emissions also contribute to the N2O enhancement ( Fig. 8 and Fig. S12). However, when we convolved the monthly mean fire contributions from the Global Fire Emissions Database (GFED, https://www.globalfiredat.org) with the surface influence footprints (as described above), we find that the wildfire produced N2O is minimal for this profile (~0.2 ppb), suggesting fires of anthropogenic or urban origin might be the 740 source of that contribution (Figs. 8A-C, 9, S12 and S13).

Conclusions
N2O mixing ratios at 1Hz were obtained during the NASA ATom airborne program by applying a new spectral retrieval The high resolution of this data set (10 s) allowed us to study the factors influencing the enhancements in the N2O tropospheric mixing ratios associated with biomass burning, human activities such as urban and industrial emissions.
The highest N2O mixing ratios are close to the Equator, extending through the tropospheric column. Episodes of the strongest N2O enhancements were observed close to the Equator and also in a number of locations at mid-latitudes. We 760 use the information given by the vertical profiles of N2O and a variety of chemical tracers together with footprints computed every 60 s along the flight track to identify and trace the sources of these N2O enhancements. N2O enhancement events were more frequent in the Atlantic than in the Pacific.
Over the Atlantic, excess N2O together with the co-occurrence of other pollutants suggested that industrial and urban N2O emissions originated in distant locations such as western and southern Africa, the Middle East, Europe and 765 South America may be significantly greater than the emissions from biomass burning in Africa. This view is supported by our observations of a strong contribution to N2O from oil and gas emissions from the Niger River Delta in Africa.
Possibly the correlations observed between N2O and SO2 (r 2 = 0.90) could be used to estimate N2O emissions from oil and gas. Over the southern Pacific Ocean and the tropical Atlantic Ocean, we observed a significant number (>12) of profiles where enhancements in N2O were associated with increased H2O2 and PAA, and notably less well correlated 775 with HCN or CO. Since H2O2 and PAA are products of photochemical pollution, this observation raised the question as to whether significant N2O may be produced by heterogenous processes involving HONO or NOx reactions in acidic aerosols close to sources, or in very heavily polluted areas. It is hard to make a definite conclusion based on measurements so far from the most active regions. Studies directed at understanding this question would have to be carried out directly in the polluted areas. Because agricultural activities do not have unique tracer signatures, we were 780 not able to distinguish contributions from cultivated and natural soils to N2O emissions from the ATom data. Previous airborne studies have observed these inputs, using flights in agricultural areas (Kort et al, 2008), and at towers in these regions (e.g., Nevison et al, 2017;Miller et al., 2008).
Our study shows that airborne campaigns such as ATom can help trace the origins of biomass burning and industrial emissions and investigate their impact on the variability of tropospheric N2O, providing unique signatures in vertical 785 profiles and with covariate tracers. We hope that the information provided by the global tropospheric N2O profiles from the ATom mission will help better diagnose and reduce uncertainties of atmospheric chemical transport models for constraining the N2O global budget.

1185
Correlations between N2O and APO, HCN, SO2, and propane between 1 and 3 km show possible contributions from marine upwelling, biomass burning and oil and gas industry, supported by the footprints (s represents the slope of the linear fit). (C) Footprint maps tracing surface regions influencing mixing ratios measured at the altitude ranges of 0-1, 2-4, 4-5, 5-7 and 7-10 km, respectively. Blue square shows the sample point. Values below 3 ppt / nmol m -2 s -1 are not included. Note that the APO axes are reversed.