HCOOH in the remote atmosphere: Constraints from Atmospheric Tomography (ATom) airborne observations

Abstract

Formic acid (HCOOH) is an important component of atmospheric acidity but its budget is poorly understood, with prior observations implying substantial missing sources. Here we combine pole-to-pole airborne observations from the Atmospheric Tomography Mission (ATom) with chemical transport model (GEOS-Chem CTM) and back trajectory analyses to provide the first global in-situ characterization of HCOOH in the remote atmosphere. ATom reveals sub-100 ppt HCOOH concentrations over most of the remote oceans, punctuated by large enhancements associated with continental outflow. Enhancements correlate with known combustion tracers and trajectory-based fire influences. The GEOS-Chem model underpredicts these in-plume HCOOH enhancements, but elsewhere we find no broad indication of a missing HCOOH source in the background free troposphere. We conclude that missing non-fire HCOOH precursors inferred previously are predominantly short-lived. We find indications of a wet scavenging underestimate in the model consistent with a positive HCOOH bias in the tropical upper troposphere. Observations reveal episodic evidence of ocean HCOOH uptake, which is well-captured by GEOS-Chem; however, despite its strong seawater undersaturation HCOOH is not consistently depleted in the remote marine boundary layer. Over fifty fire and mixed plumes were intercepted during ATom with widely varying transit times and source regions. HCOOH:CO normalized excess mixing ratios in these plumes range from 3.4 to >50 ppt/ppb CO and are often over an order of magnitude higher than expected primary emission ratios. HCOOH is thus a major reactive organic carbon reservoir in the aged plumes sampled during ATom, implying important missing pathways for in-plume HCOOH production.

TOC Graphic:

Observed formic acid (HCOOH) disribution below 5 km during ATom-3. Each data point is colored by the mean HCOOH mixing ratio for the surrounding 1.5° latitude ×1.5° longitude bin along the flight track. The 3 panels inset show the modeled (red) vs. observed (grey) median vertical profiles for example flights demonstrating the influence of fire emissions (right, flight on 05/14/2018 off the west African coast), combined fire and urban sources (middle, flight on 05/21/2018 over California), and ocean uptake (left, a tropical Pacific flight on 10/06/2017) on atmospheric HCOOH.

1. Introduction

Formic acid (HCOOH) is ubiquitous in the atmosphere, one of the largest drivers of precipitation acidity in remote regions^1–3, and a major OH sink in the aqueous phase⁴. HCOOH therefore acts as a pH buffer and affects aqueous phase chemistry through its effects on oxidant levels, pH-dependent reaction rates, and solubilities. Prior work has shown HCOOH to be present in the atmosphere at levels far greater than can be explained based on its known sources and sinks^5–8; however, such studies have mainly focused on terrestrial and near-source regions. Here we combine observations from the NASA DC-8 flying laboratory during the Atmospheric Tomography Mission (ATom-3 and 4) with a global chemical transport model (GEOS-Chem CTM) to provide the first global in-situ characterization of HCOOH in the remote atmosphere.

Recent evaluations of the global HCOOH budget have estimated its total known source at ~60 Tg/yr with subsequent removal dominated by deposition^{5, 6}. Photochemical production from volatile organic compounds (VOCs) is believed to be the dominant HCOOH source over forested regions^{9, 10}, urban areas¹¹, and on a global scale^{5, 6}, with direct emissions from the terrestrial biosphere¹², soils¹³, ants¹⁴, and urban/industrial activities¹⁵ playing more modest roles. Secondary formation pathways that have been identified include hydroxyl radical (OH)- and ozone (O₃)-initiated oxidation of isoprene, monoterpenes, and their oxidation products¹⁶, ozonolysis of other alkenes¹⁷, and alkyne oxidation¹⁸. Acetaldehyde has also been shown to be a precursor of HCOOH via its photo-tautomerization to vinyl alcohol¹⁹; subsequent work using the GEOS-Chem CTM found that this pathway can account up to 60% of the modeled HCOOH production over remote oceans²⁰. Aerosol aging has been implicated as another HCOOH source through heterogeneous (or condensed-phase) oxidation by OH²¹, ozonolysis²², and photolysis^23–26.

HCOOH is also produced from biomass burning^27–30, and recent work indicates that this source may be larger than previously thought. Yokelson et al. (2009)³¹ find that in many biomass burning plumes organic acids are the third most important carbon reservoir—behind CO₂ and CO and ahead of CH₄. The same authors report organic acid:CO enhancement ratios of up to 18% (mol/mol) in aged fire plumes—with secondary acid production exceeding the abundance of known precursors—and invoke unmeasured pyrogenic VOCs as explanation. Chaliyakunnel et al (2016)³², Pommier et al. (2017)³³, and Yamanouchi et al. (2020)³⁴ likewise observed fire HCOOH:CO enhancement ratios much higher than primary emission ratios could explain, implying substantial HCOOH production during fire plume aging.

High levels of HCOOH have also been measured in remote areas, with a variety of formation mechanisms proposed. For example, Baboukas et al. (2000)³⁵ argued that an indirect marine HCOOH source of ~50 Tg/yr globally was required to explain measured mixing ratios (averaging ~450 ppt) in the Atlantic marine boundary layer (MBL). Dibb and Arsenault (2002)³⁶ measured higher HCOOH in firn air (up to 6 ppb) than in the overlying atmosphere and invoked a source from the oxidation of organics in snow. Dibb et al. (1996)³⁷ presented strong evidence of enhanced formic acid at Summit, Greenland from the long-range transport of wildfire smoke. Jones et al. (2014)³⁸ measured HCOOH concentrations in the European Arctic MBL that were 50% higher than over the surrounding land areas, and proposed a source from stabilized Criegee intermediates following photolysis of ocean-released diiodomethane (CH₂I₂). Mungall et al. (2017)³⁹ reported HCOOH levels of up to 4 ppb in the Canadian Arctic MBL, with evidence pointing to a photochemical or heterogeneous source from the sea surface microlayer.

The above findings highlight the need for a better understanding of atmospheric HCOOH and its sources and sinks. A key open question is whether the inferred missing HCOOH source persists throughout the global atmosphere (for example, as inferred for acetaldehyde in Wang et al. (2019)⁴⁰), which would imply a ubiquitous long-lived precursor pool, or if it primarily manifests near and downwind of continental VOC source regions. The ATom campaign⁴¹ collected observations of HCOOH and a suite of related tracers from pole-to-pole over the remote Pacific and Atlantic Oceans with continuous vertical profiling and across multiple seasons (Fig. 1). The dataset thus provides a new opportunity to address the above science issues on a global scale. In this paper we apply the GEOS-Chem CTM to interpret the ATom measurements in terms of their implications for the HCOOH budget in the remote atmosphere.

Figure 1.

Observed HCOOH distribution during ATom-3 and 4. Panels a and c map the HCOOH mixing ratios measured during ATom-3 above (a) and below (c) 5 km altitude. Panels (b) and (d) show the corresponding plots for ATom-4. Each datapoint in panels a–d is colored by the mean HCOOH mixing ratio for the surrounding 1.5° latitude × 1.5° longitude bin. Panel e shows normalized frequency distributions for the ATom HCOOH observations with different linear scales at the upper range of the distribution. Some negative values occur reflecting uncertainty and variability in the instrument background signal.

2. Methods

2.1. Atmospheric Tomography (ATom) 3 and 4 aircraft campaigns

HCOOH measurements were conducted onboard the NASA DC-8 aircraft as part of the two final ATom deployments⁴¹ during September-October 2017 (ATom-3) and April-May 2018 (ATom-4). ATom was designed to (i) perform detailed chemical characterization of the global remote atmosphere, (ii) establish a comprehensive dataset of climatically and chemically important atmospheric constituents, and (iii) provide new observations to improve models⁴². Each ATom deployment spans nearly pole-to-pole over the Pacific and Atlantic Oceans with continuous vertical profiling from within the MBL to above the tropopause.

2.1.1. HCOOH measurements by Iodide Ion Time-of-Flight Chemical Ionization Mass Spectrometry (I-CIMS) during ATom-3 and 4

HCOOH was measured once per second using an iodide-ion time of flight chemical ionization mass spectrometer (I-CIMS), as described in Veres et al. (2020)⁴³. Pressure and temperature control of the sampling inlet and humidity control of the ion molecule reaction region guaranteed that the instrument response was independent of aircraft altitude and ambient conditions. The 70 cm long Teflon inlet tube was controlled to 40 °C and at ambient pressure. The pressure upstream of the ion-molecule reaction region was controlled to 140 mb, and the ion-molecule reaction region itself was controlled to 40 mb (ATom-4) or 50 mb (ATom-3) to maintain constant ion-molecule reaction times. The measurement sensitivity to HCOOH depends on water concentration in the ion molecule reaction region, and water vapor was added to this region to compensate for changing ambient humidity and maintain constant detection sensitivity. The ratio of the two reagent ions (iodide and the iodide-water cluster) served as a proxy for humidity in the instrument, and water was controlled to maintain a constant reagent-ion cluster ratio. Effects of any residual humidity changes were then corrected based on a laboratory quantification of the water-dependent HCOOH sensitivity as a linear function of the iodide:iodide-water ion ratio.

HCOOH was detected as a cluster with iodide at m/z 172.9105, with a resolution of m/Δm=5000. The instrument sensitivity to HCOOH was determined before and after the campaign by laboratory-based standard addition of HCOOH from a permeation tube source with the instrument in its flight configuration. The permeation tube was held in an oven controlled to 40 °C with a 50 sccm continuous flow of N₂. The HCOOH emission from the permeation tube was determined by catalytically converting HCOOH to CO₂ followed by CO₂ detection.⁴⁴ The 47 ng/min HCOOH permeation tube emission was mixed with the 6 slm inlet flow to give a 3.8 ppb calibration mixing ratio. The uncertainty in this calibration process is approximately 15%. Under typical operating conditions, the instrument sensitivity was approximately 8 ion counts/second/ppt HCOOH, with reagent ion signals of 3 MHz for iodide and 1.5 MHz for the iodide-water cluster. In-flight calibrations of chlorine and nitric acid demonstrated that the reagent-ion-normalized instrument sensitivity did not change over the course of each campaign.

The instrument zeros were determined in-flight for 30 s every 8 min by overfilling the inlet with ambient air drawn through a scrubber. After ATom-4, an additional background measurement capability was added that involved the periodic addition of dry N₂ at the inlet tip without going through a scrubber. These two zeroing methods were found to agree to within the variability of the background determinations. For ATom, average background signals were typically 300 ppt; these were interpolated between consecutive determinations and subtracted from the total signal to determine the ambient HCOOH contribution. Background signals varied by 50 ppt to 100 ppt between successive quantifications, imposing an additional 50–100 ppt measurement uncertainty (and a lower limit of detection) that is not reduced by averaging over times less than background measurement interval. The measurement precision was better than the instrument background uncertainty, and was typically below 10 ppt for 1 s measurements, ranging from 3–15 ppt.

2.1.2. Other measurements

The DC-8 aircraft featured a large instrument payload during ATom, and we employ the following measurements to aid interpretation of the HCOOH dataset. For species measured by more than one instrument, dataset selection was based on time response, data availability, and nominal accuracy. Carbon monoxide (CO) was measured with the NOAA Picarro cavity ring-down spectrometer (G2401-m)⁴⁵ and with the Harvard Quantum Cascade Laser System (QCLS)⁴⁶. We use here the merged CO.X product⁴¹, which is based primarily on the QCLS observations with gaps filled using the NOAA Picarro observations. We use hydrogen cyanide (HCN) observations from the Caltech chemical ionization mass spectrometer using CF₃O⁻ reagent ions (CIT-CIMS)^{47, 48}, and formaldehyde (HCHO) observations from the NASA In Situ Airborne Formaldehyde (ISAF) instrument^{49, 50}. VOC measurements used here were obtained with the NCAR Trace Organic Gas Analyzer (TOGA)⁵¹ and the UC Irvine Whole Air Sampling System (WAS)⁵². NO_x (NO+NO₂) was measured by the NOAA Nitrogen Oxides and Ozone (NO_yO₃) instrument⁵³. Black carbon (BC) was measured with the NOAA Single Particle Soot Photometer (SP2)⁵⁴. Submicron organic aerosol (OA) was measured with the University of Colorado High-Resolution Time-of-Flight Aerosol Mass Spectrometer (HR-AMS)^{55, 56}. BC and OA are both reported for T = 273 K and P = 1013 hPa; we perform all unit conversions accordingly. Submicron organic carbon aerosol (OC) is derived from OA based on the OA:OC ratio reported by the instrument team. We use the Particle Analysis by Laser Mass Spectrometry (PALMS) biomass burning particle number fraction⁵⁷ in our analysis to aid in plume source attribution. Finally, we use the cloud indicator values from the Cloud, Aerosol, and Precipitation Spectrometer (CAPS)⁵⁸ to identify within-cloud sampling periods.

2.2. GEOS-Chem simulation

We use the GEOS-Chem CTM in its Classic mode⁵⁹, with detailed HO_x-NO_x-VOC-ozone-halogen-aerosol chemistry following JPL/IUPAC recommendations and photolysis frequencies calculated using Fast-JX algorithms^60–62. The simulation is driven by assimilated meteorological data (Goddard Earth Observation System Forward Processing, GEOS-FP) from the NASA Global Modeling and Assimilation Office (GMAO) and conducted at 2°×2.5° horizontal resolution with time steps of 15 min (transport/convection) and 30 min (chemistry/emission). We use the TPCORE advection algorithm⁶³, convective mass fluxes from the meteorological archive⁶⁴, and non-local boundary layer mixing⁶⁵.

We employ a customized GEOS-Chem simulation for HCOOH building on prior work by Shaw et al. (2018)²⁰, Millet et al. (2015)⁵, and Paulot et al. (2011)⁶. Photochemical HCOOH production proceeds as described by Millet et al. (2015), with photo-tautomerization of acetaldehyde updated according to Shaw et al. (2018)²⁰. Global anthropogenic emissions for HCOOH and other species are from the Community Emissions Data System inventory⁶⁶, overwritten for Africa by the Diffuse and Inefficient Combustion Emissions in Africa dataset (DICE-Africa)⁶⁷. Global HCOOH emissions from agriculture and natural soils are scaled to the corresponding ammonia and NO_x fluxes following prior work⁶. Direct HCOOH biogenic emissions are calculated using the Model of Emissions of Gases and Aerosols from Nature version 2.1 (MEGAN v2.1)⁶⁸, implemented in GEOS-Chem as described by Hu et al. (2015)⁶⁹. Biomass burning emissions are based on the Global Fire Emissions Database with small fires (GFED4s)⁷⁰ with recommended species-specific emission factors (http://www.globalfiredata.org)⁷¹. Ocean emissions of HCOOH precursors (alkenes and isoprene) are estimated following Millet et al. (2015)⁵ and Paulot et al. (2011)⁶. Dry deposition of HCOOH and other species is computed using a modified Wesely scheme^{72, 73}; over oceans this is equivalent to an infinite liquid-phase sink and hence unidirectional air-to-sea flux. The resulting monthly-mean HCOOH dry deposition velocities during the timeframes of ATom-3 and ATom-4 typically range between near-zero and 0.8 cm/s over land and reach 2.2 cm/s over ocean. Wet deposition is as described by Amos et al. (2012)⁷⁴ and Mari et al. (2000)⁷⁵. Uptake of HCOOH on ice and dust follows Paulot et al. (2011)⁶.

2.3. Back trajectory analysis

We use atmospheric back trajectories computed along the ATom flight tracks with the Traj3D model^{76, 77} to interpret the observations in terms of their source footprint and source/sink influences. The model is driven by National Centers for Environmental Prediction (NCEP) Global Forecast System (GFS) 0.5°×0.5° resolution meteorology. For each minute along the flight track, a cluster of 245 trajectories is initialized in a 0.3°latitude × 0.3° longitude × 20 hPa cube centered on the aircraft position. Each trajectory is then run 30 days backwards with a 3-hour time-step. Traj3D model results used here include individual trajectories, cluster ensemble distributions, and cluster-average results.

Two influence processes are employed in this work, as follows. (1) Fire influence. A trajectory is considered fire-influenced if it encountered a fire hotspot at a time and location indicated by one of the following three products: Moderate Resolution Imaging Spectroradiometer (MODIS) Fire Radiative Power (FRP) (Collection 6)⁷⁸; Visible Infrared Imaging Radiometer Suite (VIIRS) 375-m FRP⁷⁹; or Global Fire Emissions Database (GFED) v2. Fire plume heights were calculated based on FRP for MODIS and VIIRS, and assumed to be within the boundary layer for GFED. (2) Land boundary layer (BL). A trajectory is considered to be influenced by the land boundary layer if it entered the BL over any continent.

3. Results and Discussion

3.1. HCOOH distribution over remote oceans and role of fire plumes in driving variability

Figure 1 shows horizontal and frequency distributions of HCOOH mixing ratios observed during ATom-3 and 4. We see that the ATom flights sampled a broad range of HCOOH concentrations from below the instrumental LOD to above 30 ppb (Fig. 1f). However, the majority of the distribution is characterized by very low concentrations, with 80% (ATom-3) and 61% (ATom-4) of the data falling below 100 ppt (Fig. 1). While the instrumental LOD prevents accurate quantification of the lowest observed concentrations, the data clearly reveal sub-100 ppt HCOOH background concentrations over most of the remote oceans. More frequent HCOOH enhancements are apparent during ATom-4 than ATom-3 (Fig. 1e–f); as will be seen this reflects strong fire influence during this campaign. In general, lower HCOOH mixing ratios are seen aloft (Figs. 1a/1b vs. 1c/1d), but with clear enhancements in certain regions: east of Australia, east of South America, and over the springtime northern oceans.

Figure 2a compares the vertical HCOOH distribution observed during ATom to that seen in prior aircraft campaigns (Table S1) over the ocean. We see that the median ATom HCOOH profile is very similar to that observed during SONEX (10–11/1997)⁸⁰ and PEM-Tropics A (08–09/1996)⁸¹, particularly in the free troposphere. PEM-West A (09–10/1991)⁸², PEM-West B (02–03/1994)⁸³, and INTEX-B (04–05/2006)^{84, 85} sampled closer to land, with the associated Asian/North American outflow leading to higher median HCOOH concentrations. TRACE-A (09–10/1992)⁸⁶ encountered extremely high HCOOH levels in the south Atlantic due to biomass burning in South America and southern Africa. Together, these campaigns are consistent in revealing HCOOH concentrations on the order of <100 ppt over the remote oceans, rising to >100 ppt closer to continental regions, and much higher under strong fire influence.

Figure 2.

Median HCOOH vertical profiles measured during ATom-3 and 4, and comparison with (a) prior aircraft observations over ocean and (b) GEOS-Chem model predictions. The vertical bin resolution differs between panels (a) and (b) (a: 2 km; b: 0.5 km below 3 km altitude, 1 km above 3 km altitude) to reflect differing sampling densities between the campaigns. Horizontal bars indicate 25th-75th percentiles for each vertical bin. For panel (a) observations over land have been filtered out, and the SONEX dataset further employs ozone and latitude filters following Singh et al. (2000)¹³⁴. See Table S1 for details on the aircraft campaigns plotted in panel (a).

Figure 2b compares the ATom-3 and 4 vertical HCOOH profiles with those simulated by GEOS-Chem. For both campaigns the model significantly underestimates the median observed concentrations in the lower atmosphere, with the discrepancy more severe during ATom-4. On the other hand, the model captures the median HCOOH concentrations seen in the mid-to-upper free troposphere with no appreciable bias. This contrasting behavior will be further discussed in Sect. 3.2.

Figure 3 compares interpolated (0.05° latitude × 250 m) latitude curtains for HCOOH, CO, BC, and HCN based on the ATom-3 and 4 observations. While CO exhibits a clear latitudinal gradient with elevated concentrations throughout the northern high latitudes, no consistent gradient is detectable for HCOOH. Instead, we observe persistently low (< 100 ppt) background HCOOH concentrations throughout most of the remote atmosphere with discrete enhancements at specific locations associated with terrestrial outflow. The strongest such enhancements occur during spring (Southern Hemisphere for ATom-3; Northern Hemisphere for ATom-4) and consistently over the tropical Atlantic (Fig. 3). This pattern is very similar to that seen for BC, a known combustion tracer. In fact, the HCOOH correlation with BC is higher than with any other ATom-measured species (R = 0.72 and 0.74 for the entire ATom-3 and 4 datasets, respectively; flight-specific values reach 0.86; Fig. S1). The localized nature of the strongest enhancements, their locations and their distance from population centers, together point to a predominant role for biomass burning in driving remote HCOOH variability.

Figure 3.

Interpolated tracer latitude-altitude cross-sections based on ATom-3 (Northern Hemisphere fall) and 4 (Northern Hemisphere spring) in-situ observations. 1 Hz data are gridded and interpolated into 0.05° latitude × 250m altitude bins.

BC is also emitted from other form of combustion, and the Figure 3 enhancements presumably also include some contributions from non-fire sources such as urban pollution. Notably, the HCOOH correlation with HCN (a biomass burning tracer) is not as high as with BC (Fig. S1); this may reflect the large variability in fire HCN emissions⁸⁷ as well as the differing removal pathways for HCN and HCOOH. Overall, the strong importance of fires for HCOOH is confirmed by its tight correlation with the trajectory-derived fire influence (Fig. S2)—HCOOH enhancements under recent fire influence are an order of magnitude higher than under general land boundary layer influence. Later (Sect. 3.4.2) we assess the importance of primary emissions versus in-plume production in driving these HCOOH enhancements, and explore the role of HCOOH as a reactive carbon reservoir in highly aged fire plumes.

Despite the high overall HCOOH-BC correlation, we see in Figure 3 instances where their spatial patterns differ. For example, elevated HCOOH is observed in the MBL at ~30° N over the Pacific during ATom-3 with no accompanying BC enhancement. Also during ATom-3, comparable HCOOH enhancements are seen over the tropical and southern Atlantic, whereas for BC the tropical enhancements are far more pronounced. Such differences can arise from factors such as secondary production of HCOOH, non-pyrogenic sources, and differing removal rates.

The following sections combine the ATom observations with GEOS-Chem model predictions to i) evaluate current model treatment of HCOOH in the remote marine atmosphere (outside of fire plumes) in terms of evidence for broad-scale missing sources, and ii) diagnose signatures of HCOOH removal and the degree to which these are captured in the GEOS-Chem simulation. We return to examine the role of fires for remote HCOOH in Section 3.4.

3.2. Is there a ubiquitous missing HCOOH source in the background free troposphere?

Several prior studies have demonstrated the existence of large missing HCOOH sources^{5–8, 16, 39}, which is consistent with the model-measurement discrepancies seen during ATom in the lower troposphere (Fig. 2b). On the other hand, Figure 2b shows that for both ATom-3 and ATom-4 GEOS-Chem is able to capture the median observed HCOOH abundance in the free troposphere with no detectable low bias. Median model levels in the free troposphere are generally in the 50–100 ppt range; if there were persistent missing sources leading to concentrations significantly >100 ppt they would be detectable given the instrumental LOD. Strong model HCOOH underestimates are seen in the vicinity of plumes and in the springtime northern latitudes, but do not occur ubiquitously across the remote free troposphere (Fig. 4) or at any altitude when >10 days removed from any land boundary layer or fire influence (Fig. S2). In fact, there is some indication of a positive model bias in the free troposphere (e.g., at high altitudes over the tropical Atlantic; Fig. 4) which we discuss further in Section 3.3.

Figure 4.

(a–c): Median gridded (10° latitude × 500m altitude) HCOOH cross-sections. Plotted are the (a) ATom observations, (b) GEOS-Chem model predictions, and (c) absolute model bias. (d): Kernel-smoothed probability density distributions and box-whisker plots (5th-25th-50th-75th-95th percentiles) for the HCOOH:CO ratio as measured during ATom (black) and modeled by GEOS-Chem (red) with numbers inset representing (left to right) the 5th, 50th, and 95th percentiles of the distribution.

The comparisons above imply that the large missing sources of HCOOH inferred in prior work are mainly terrestrial and sufficiently short-lived that their effects do not manifest as a consistent detectable model bias in the ATom dataset. Figure S3 shows that median levels for key measured HCOOH precursors are indeed very low during ATom, and this is generally captured by the model (though with underestimates for ethyne and acetaldehyde). We likewise conclude that the fire plume enhancements discussed earlier do not reflect a source of sufficient global magnitude to yield a detectable model HCOOH bias throughout the remote atmosphere.

Recent ATom-based analyses by Wang et al. (2019)⁴⁰ and Travis et al. (2020)⁸⁸ point to a widespread missing acetaldehyde source in the free troposphere over remote oceans, which is also seen here (Fig. S3). Unidentified precursor VOCs are one possible explanation; if so, it appears that those VOCs do not give rise to large amounts of HCOOH. In addition, while acetaldehyde itself is a HCOOH precursor via tautomerization to vinyl alcohol^{19, 89–92}, we find that the model bias for this species is insufficient to have a large broad-scale impact on HCOOH. Specifically, a sensitivity simulation doubling vinyl alcohol source from acetaldehyde yields a modeled HCOOH increase of <1 ppt over most remote areas. Thus, while acetaldehyde tautomerization can be a large fraction of the total modeled HCOOH production rate in remote areas²⁰, the absolute magnitude is then relatively small.

3.3. Air-sea exchange and wet scavenging of HCOOH

HCOOH has a high gas-to-aqueous partitioning tendency with an intrinsic Henry’s law constant of k_H = 8917 M/atm at 298K^93–95. Its aqueous uptake is further enhanced by dissociation (pKa = 3.75) in natural solvents such as rainwater (pH ~ 5.6; $k_H^{eff}=6\times {10}^5$ M/atm at T = 298 K) and ocean water (pH ~ 8.1; $k_H^{eff}=2\times {10}^8$ M/atm at T = 298 K). As a result, wet scavenging is thought to be the dominant global sink of atmospheric HCOOH⁶, and the signature of this process is indeed evident in the ATom dataset. For example, HCN and CO co-enhancements in the tropical Atlantic upper troposphere (Fig. 3) indicate the presence of transported fire plumes. However, no comparably strong HCOOH or BC enhancements are apparent in the same locations, implying wet removal of these species by convective updrafts, clouds, or precipitation during plume transport. The differing behavior for HCOOH versus HCN can be explained by the much weaker solubility (k_H = 12 M/atm at 298K⁹³) and acid dissociation (pKa = 9.2) for HCN. In the GEOS-Chem simulation we thus derive a global HCOOH lifetime of ~8 days against wet deposition, whereas this is a negligible removal pathway for HCN^{96, 97}. The fact that the model frequently overestimates HCOOH at high altitude, particularly over the tropical Atlantic (Fig. 4; Section 3.2), may reflect an underestimate of convective HCOOH removal–as has been identified previously in the model for black carbon and sea salt aerosols⁹⁸.

The ATom data also include segments with clear evidence of oceanic HCOOH uptake. Figure 5 shows measurements of HCOOH and related tracers during two flights with minimal continental influence (based on CO, BC, and HCN) over the Tropical Pacific and Antarctic Oceans. We see strong HCOOH depletion in the MBL for nearly all of the vertical profiling maneuvers (others are apparently influenced by fire or pollution plumes). HCN is likewise depleted in the MBL, whereas CO is enhanced. MBL depletion of HCOOH could in theory be explained either by cloud scavenging or by ocean uptake. However, the timeseries in Figure 5 show that MBL depletion of HCOOH is not dependent on cloudy conditions—arguing against cloud scavenging as the major driver. Similar HCOOH depletion in the MBL was previously observed during an airborne ascent/descent over the European Arctic ocean³⁸.

Figure 5.

Ocean uptake of HCOOH as evident during two background flights: (a) a tropical Pacific flight on 10/06/2017, and (b) an Antarctic flight on 05/09/2018. Left panels show timeseries for three trace gases (black) and altitude (blue). Shaded regions (grey) in the HCOOH panel indicate cloudy conditions (liquid-phase, mixed-phase, or cirrus clouds) based on CAPS observations. Right panels show the corresponding median observed vertical profiles for the entire flight (in black) with horizontal bars denoting the 25th–75th percentiles. The observed HCOOH profile is compared to GEOS-Chem model predictions (in red). Panel (c) shows the two flight tracks (black).

Figure 5 also compares the HCOOH vertical profiles from the above remote flights with model predictions, showing that GEOS-Chem generally reproduces the observed 0–5 km vertical gradient. While the observations have high relative uncertainty at these low concentrations, this agreement supports the idea that an air-side resistance-in-series deposition mechanism can adequately represent ocean HCOOH uptake without considering its degree of saturation or liquid-side resistance. For example, based on the effective Henry’s constant above, 1 ppb and 1 ppt of atmospheric HCOOH correspond at equilibrium to surface ocean [HCOOH_(aq) + HCOO⁻] concentrations of 0.2 M and 0.2 mM, respectively (and higher at lower temperatures), with formate as the predominant dissolved form ([HCOOH_(aq)]:[HCOO⁻] = 4.5×10⁻⁵ at pH = 8.1). With very few exceptions⁹⁹, literature-reported seawater formate concentrations are in the μM or nM range (Table S2^100–104)—orders of magnitude below the saturation concentration. It is clear therefore that the oceans are strongly undersaturated for HCOOH and that the bulk ocean-atmosphere flux should be effectively unidirectional. In this way we find in GEOS-Chem that ocean uptake accounts for 14% of global annual HCOOH removal through dry deposition, corresponding to 5% of the total sink (Table S3).

However, the MBL depletion of HCOOH discussed above is not consistently present across all of the ATom flights and profiles. In particular, the mean ATom-3 and ATom-4 latitude-altitude cross-sections (Fig. 3) show consistent MBL depletion for HCN but not for HCOOH. Based on solubility alone we would conversely expect a stronger uptake signal for HCOOH than for HCN; the lack of a consistent HCOOH ocean uptake signature in the ATom observations implies that other processes are obscuring that signal. For example, Figure S4 shows a series of ATom vertical profiles in which HCN is characteristically depleted in the MBL but HCOOH is strongly enhanced. BC and CO co-enhancements for these same profiles suggest fire and/or pollution influence. However, given their expected co-emission and solubility difference it is then surprising that HCN would be scrubbed from the plume while the HCOOH enhancement persists. This may reflect ongoing secondary production during plume transport, or a marine boundary layer source of uncertain mechanism^{38, 39}.

3.4. A large and variable HCOOH sources from fires

We have seen above that fire plumes are key drivers of HCOOH variability in the remote atmosphere sampled by ATom. This influence is strongly underestimated by GEOS-Chem. Figure 4 plots the latitude-altitude HCOOH curtains from Figure 3 as median values over 10° latitude × 500 m altitude bins for comparison with the model. The enhancement patterns visible in Figure 3 still emerge in the median values and at the coarser spatial resolution of Figure 4. While the model exhibits some fidelity in the simulated enhancement locations, predicted magnitudes are far weaker than observed. This tendency is also seen in the HCOOH:CO concentration ratios: Figure 4 shows that during ATom-3 and the Pacific leg of ATom-4 the model captures the central part of the HCOOH:CO distribution but misses the upper tail. A larger model bias is evident over the Atlantic during ATom-4, where the model underestimates the median HCOOH:CO ratio by a factor of >2 and the 95th percentile by a factor of >4.

We proceed to examine these HCOOH enhancements in terms of potential reasons for the model underpredictions, employing as comparison metric the normalized excess mixing ratios (NEMR) relative to CO. The NEMR is commonly used to quantify dilution-corrected above-background plume enhancements: at-source the NEMR corresponds to a molar emission ratio (ER), while downwind values also reflect ensuing production and loss. We performed a case-by-case identification of plumes intercepted during ATom meeting any two of the following criteria: (a) peak HCOOH > 1 ppb, (b) peak BC > 100 ng/m³, (c) peak CO > 175 ppb (using one-second data in all cases). In this way, 53 plume intercepts were identified during ATom-3 and 4 ranging from <1 min to >20 min duration. Plume identification, NEMR calculation, and source attribution are described for each intercept in the Supporting Information (Table S4, Sect. S2). In particular, we find that 38 of the 53 identified plume intercepts exhibit clear biomass burning signatures and are thus treated as fire plumes in the analyses that follow. The remaining 15 are likely to be partly or mainly anthropogenic.

HCOOH enhancements across all 53 plume intercepts vary from ~1 to >30 ppb, with corresponding CO enhancements of 9–270 ppb. The resulting HCOOH NEMRs (calculated for each plume intercept as the reduced major axis HCOOH:CO slope; Table S4 and Fig. 6) range from 3.4 ppt/ppb CO in the northern Pacific (plume intercept #19, Sect. S2) to >200 ppt/ppb CO in plumes with potential urban/anthropogenic influence (plume intercepts #50–53, Sect. S2).

Figure 6.

HCOOH normalized excess mixing ratios (NEMRs) relative to CO as a function of transport time for ATom plumes, and comparison to prior studies. Transport times (ND: not determined) are estimated based on back trajectory analysis as described in-text. Colored dots indicate fire plumes and plumes with anthropogenic influence (grey). Plumes marked with an “×” denote encounters after a mid-flight instrumental change on 05/14/2018 (see text). Also shown are primary HCOOH emission ratios (ER) as compiled by Andreae (2019)¹²⁵ and Akagi et al. (2011)⁷¹, and HCOOH NEMR estimates inferred from prior atmospheric observations. These include: aircraft studies featuring downwind plume sampling (up to 4.5 hours aging)^{27, 31, 105, 121}, ground-based remote sensing by Fourier transform infrared spectrometry (g-FTIR)^{34, 135–137}, and space-based observations from the ACE-FTS^{29, 30, 123, 124, 138, 139}, and TES/IASI satellite instruments^{32, 33, 140}.

3.4.1. Case study: Large HCOOH enhancements in African plumes

We find that 19 of the 53 identified plume intercepts are due to African fires (with 15 from northwest Africa and 4 from southern Africa; Sect. S2.1). Of these, plume intercept #40 off the African coast (~16 UTC on 05/14/2018; Fig. 7a) featured HCOOH mixing ratios exceeding 15–30 ppb—the highest values seen across all of ATom. Highly enhanced levels of CO (~350 ppb), HCN (~1.5 ppb), and BC (~2 μg/m³) were also observed. Back-trajectory analysis indicates that the above plume underwent >2 days transport after emission in northwest Africa before being intercepted by the aircraft (Fig. 7c and Sect. S2.1). Explaining the observed in-plume HCOOH mixing ratios requires i) sufficiently large primary HCOOH emissions that 15–30 ppb still remained after >2 days dilution, ii) strong secondary HCOOH production during plume aging, or iii) some combination of the two. HCOOH enhancements this large have only been reported once previously: for a fresh grass fire plume in Southern Africa¹⁰⁵. In that case the HCOOH enhancement dropped to 5 ppb after 30 min and to below the detection limit after 40 min. Most other in-situ studies (even those conducted near-source) have reported HCOOH mixing ratios below 12 ppb^{5, 9, 106–111}.

Figure 7.

Aged west African fire plumes sampled on 05/14/2018. Panels (a) and (b) show HCOOH, HCN, CO, and BC time series for the entire flight, with 7 plume intercepts (#37–43) marked inset. An additional plume (between #38 and #39) was encountered during an HCOOH instrumental zero and is therefore omitted. Panel (c) shows HYSPLIT back trajectories (red, 1-day interval marked) for 3 example plume intercepts. Panels (d)–(k) show correlations between select pyrogenic species and CO for all plume intercepts. The resulting NEMR range (obtained from the regression slopes for the 7 plume intercepts) is compared to expected primary emission ratios (ER, shaded in grey) for that species based on compiled data ranges for relevant fuel types (forests, savanna and grassland, and agricultural residue¹²⁵). For TOGA-measured species only plume intercepts including 3 or more data points (#37, #38, #40, #41) are retained for NEMR analysis. Convective influence is negligible for these plumes based on back trajectory analysis. The lifetime (unit: days) is shown for each species with respect to its dominant loss pathway^{5, 141–144}. An OH concentration of 2×10⁶ molecule/cm³ is used to calculate nominal lifetimes to OH oxidation^{145, 146}.

The same 05/14/2018 ATom-4 flight included a total of 8 plume intercepts identified as fire outflow from west Africa, and HCOOH:CO NEMR values were quantified for 7 of these (the 8^th occurred during an instrument zeroing procedure). Figure 7 shows that HCOOH exhibits quite different behavior compared to other species sampled across these plumes. HCN and BC, primary fire tracers that are unreactive towards OH, exhibit similar NEMRs across the 7 plume intercepts that are within the range of literature-reported ERs for relevant fire types (savanna, tropical forest, agriculture). Benzene, a primary combustion tracer that is more photochemically reactive (τ=4.8 days at OH=2×10⁶ molecules/cm³ and k_benzene+OH=1.2×10⁻¹² cm³/molecule/s¹¹²) exhibits NEMRs 44–78% lower than the expected primary ERs—consistent with a few days of aging. Acetaldehyde likewise exhibits NEMR values lower than its primary ERs, reflecting the net effect of direct emissions, secondary production, and a short photochemical lifetime. NEMR values for acetone, on the other hand, are 1.4–4.2× higher than the corresponding ERs and indicate net secondary production during plume aging^{113, 114}. Methanol, despite having a photochemical lifetime similar to that of benzene, exhibits NEMR values consistent with its primary ERs, likely reflecting in-plume secondary production offsetting chemical loss^115–117—as also inferred from the ATom observations by Bates et al. (2021)¹¹⁸. The I-CIMS instrument underwent an operational change at ~14:30 UTC during this flight which may explain the HCOOH NEMR bifurcation seen in Figure 7. However, even if we take the conservative assumption that the post-14:30 HCOOH NEMRs on this flight should be reduced by ~2× to match the earlier values, Figure 7 shows that HCOOH is unique among the plotted tracers in exhibiting in-plume NEMRs 1–2 orders of magnitude higher than expected primary ERs—suggesting substantial secondary production en-route.

3.4.2. HCOOH NEMR variability and relationship with plume age

HCOOH NEMRs vary significantly across the ATom plumes investigated here. Values range from a low of 3.4 ppt/ppb CO in a high-altitude fire plume over the North Pacific (plume intercept #19, flight 20171027; the same plume featured the highest HCN mixing ratios encountered during ATom) to >50 ppt/ppb CO. Particularly high NEMR values were observed in a number of plumes that appear to contain a mixture of biomass burning and urban/anthropogenic contributions (Sect. S2.2 and Fig. 6). These large HCOOH enhancements (e.g., >200 ppt/ppb CO for California plumes encountered near the end of ATom-4) may reflect accelerated oxidation of VOC-rich biomass burning emissions in the presence of ample anthropogenic NO_x^{119, 120}.

The highest HCOOH NEMRs seen during ATom and highlighted above exceed any values reported previously in fire emissions or urban plumes. However, HCOOH has not been routinely measured in airborne fire campaigns, and prior studies including HCOOH measurements have focused on relatively fresh plumes (< 4.5 hrs) without characterizing their longer-term evolution. For example, Goode et al. (2000)¹²¹ studied the evolution of two Alaskan forest fire plumes and found that the HCOOH NEMR doubled (from 6–8 to 12–16 ppt/ppb CO) after 2–3 hrs evolution. An immediate increase of the HCOOH NEMR (to 10 ppt/ppb CO) was observed in the first 20 min after emission from a savanna fire in southern Africa, but this was soon followed by a decrease to below the instrumental detection limit¹⁰⁵. During MILAGRO, the HCOOH NEMR in a tropical fire plume was observed to increase at a rate of 0.0033/hr (from 2.5 to ~5 ppt/ppb CO) over the first 1.5 hr of aging³¹. Akagi et al. (2012)²⁷ report for a prescribed chaparral fire plume in California that the HCOOH NEMR increased by a factor of 7 (from 0.7 to 4.8 ppt/ppb CO) over 4–4.5 hrs aging. Finally, Yokelson et al. (2009)³¹ report an even higher NEMR for total organic acids of up to 180 ppt C/ppb CO in aged smoke, but do not specify the fraction due to HCOOH. In the case of urban/anthropogenic sources¹¹, direct HCOOH emissions have been found to be relatively small (~ 1 ppt/ppb CO^{106, 109, 110}) with subsequent secondary production yielding enhancements of up to ~20 ppt/ppb CO¹²².

Enhanced HCOOH levels in fire plumes have also been quantified using remote sensing observations. Tereszchuk et al. (2011, 2013)^{123, 124} combined ACE-FTS data with trajectory modeling to examine the evolution of pyrogenic trace species in biomass burning plumes from a variety of geographic regions and found a net decrease of the HCOOH NEMR over 7 days aging. More recently, satellite data from IASI³³ and TES³² have been used to study the HCOOH source from key biomass burning regions, with the latter study inferring substantially higher enhancement ratios than expected solely from direct emissions—indicating significant secondary HCOOH production in fire plumes.

One hypothesis to explain the variable HCOOH enhancements observed during ATom is that longer plume transport times and hence aging lead to additional net HCOOH production. We test this idea by estimating physical plume ages for the identified fire intercepts. Ages are computed for each minute of the intercept window as the time since most recent fire influence based on back-trajectory analysis (Methods; Sect. S2). We find in this way that the 53 identified plume intercepts can be categorized as follows: (i) one single unambiguous transport time; (ii) a unimodal transport time distribution; (iii) a multimodal transport time distribution; (iv) variable trajectory cluster results across the plume intercept window; (v) low fire influence probability based on the back trajectories and therefore an undetermined age. We then obtain a best-estimate transport time for each plume intercept from the ensemble of trajectory results across the intercept. Finally, we group the results into plume age categories of (a) <1 day, (b) 1–5 days, (c) 5–10 days, (d) >10 days, or (e) a mixture of two neighboring classes within a-d. More details on this classification process are provided in Section S2.

Figure 6 shows the resulting HCOOH:CO plume-age dependence for the ATom dataset in relation to prior NEMR observations and direct pyrogenic emission ratios. We see that the strongest enhancements (>100 ppt/ppb CO) manifest for aged and anthropogenically-influenced plumes (with their likely faster chemistry). Conceptually, we expect increasing NEMRs during early plume aging if the rate of secondary HCOOH production exceeds that of CO, with a subsequent decrease as the plume age approaches the HCOOH photochemical lifetime and precursor VOCs become depleted. The mean trend in Figure 6 does not contradict this expectation, but the main emergent feature is that of highly variable HCOOH enhancements across plumes, with NEMRs spanning a factor of 50 or more within individual age classes. While our trajectory-based plume age estimation and the instrumental calibration for HCOOH (Sect. 2.1.1) both have some inherent uncertainty, it appears that this remaining variability must mainly reflect other factors such as source type and burn conditions that affect the HCOOH ER and subsequent plume chemistry.

3.4.3. HCOOH as a major C reservoir in aged plumes

Figure 8 presents a global view of the gas-phase reactive organic carbon (ROC) composition of the plumes encountered during ATom-3 and 4. We define ROC here to include all VOCs measured during ATom with sufficient in-plume sampling coverage to determine regression-based NEMRs, plus submicron aerosol OC (Sect. S3). We see in the map plumes spanning a wide range of intercept locations, source regions, and transit times, including: transported Eurasian fire plumes sampled over the Arctic, Alaska, and the north Pacific; plumes from the northwestern US/southwestern Canada sampled over California; Central American fire plumes sampled over the equatorial Pacific; plumes downwind of Australia, South America, and Africa; and near-source plumes sampled over California and Fiji.

Figure 8.

Gas-phase reactive organic carbon (gROC) composition in ATom-3 and ATom-4 fire plumes. Grey shaded plume intercept numbers are plumes with anthropogenic influence. Identified plumes (filled red circles) are denoted along the flight tracks (orange lines) with associated back trajectory clusters colored by the time since most recent fire influence. Plumes and trajectories shown are restricted to those with identified fire influence within the previous 10 days. Pie charts show in-plume gROC speciation on an NEMR basis (ppt C/ppb CO). VOCs that are measured only with the WAS system are omitted here for consistency across plumes; Figure S5 includes such species when available. See Figure S6 for a version of this figure that includes submicron aerosol OC.

The pie charts in Figure 8 quantify the gas-phase ROC (gROC) composition for selected sampled plumes on an NEMR basis (ppt C/ppb CO). The 21 plume intercepts shown (of 53 total) reflect those with sampling coverage for the I-CIMS, CF₃O⁻ CIMS, ISAF, and TOGA instruments (see Methods). VOCs measured only by WAS (with its lower temporal resolution) are excluded to prevent species coverage differences between plumes. Figure S5 shows that including these species (when available) changes the HCOOH contribution by 1–26% on an NEMR basis.

We see for the plumes plotted in Figure 8 that HCOOH, with NEMRs ranging from 11–129 ppt/ppb CO, accounts for between 6 and 60% of the total detected gROC enhancements. This fractional contribution depends on measurement coverage, which is incomplete as ATom was not explicitly a biomass-burning focused mission. Based on prior estimates for the importance of unmeasured/unidentified pyrogenic VOCs⁷¹, the HCOOH fractions in Fig. 8 (and S5) could be 2–3× lower if all species were accounted for.

The lowest HCOOH contributions (6–23%) are seen for relatively near-source plumes over California and Fiji, contrasting with >30% contributions for several transported Asian and African plumes. Figure S6 shows that when we expand the analysis to include both gas and particle-phase OC, OC is frequently the largest component of the in-plume ROC enhancement. However, HCOOH still typically contributes between 4 and 33% of the total gas + particle ROC enhancements shown in Figure S6. HCOOH therefore represents a surprisingly large fraction of the detected ROC pool in these plumes, despite being only a small component of primary biomass burning and anthropogenic ROC emissions. As a result, when not subject to wet deposition, HCOOH can be an important pathway for transporting combustion-emitted reactive carbon to the remote atmosphere.

3.4.4. Implied in-plume HCOOH production greater than predicted from known precursors

The fact that the HCOOH enhancements observed during ATom are, in most cases, far larger than can be explained solely based on its direct pyrogenic emissions (Fig. 6) points to secondary production as the dominant HCOOH source in these plumes. The total summed direct ER for all identified gas+particle-phase ROC components is estimated to range from ~300–500 ppt C/ppb CO across forest, savanna/grassland/shrubland, and cropland fires, which are the predominant global fire types (Table S6)^{71, 125, 126}. This increases to ~500–1200 ppt C/ppb CO when also including estimated emissions of unidentified/unmeasured ROC components⁷¹. The median HCOOH enhancement observed across the ATom fire plumes of 36 ppt/ppb CO therefore corresponds to approximately 3–7% of the total expected ROC emission from fires on an NEMR basis.

However, only a fraction of fire-emitted ROC species have been identified as HCOOH precursors, including alkenes (ethene, propene, butene, isoprene, monoterpenes), ethyne, several OVOCs (MVK, MACR, acetaldehyde), hydroxyacetone, hydroxyacetaldehyde, and submicron aerosol OC (Table S6)^{17, 18, 21, 23, 127–130}. The summed direct ER for these known gas-phase HCOOH precursors is estimated to range from 91 to 189 ppt C/ppb CO for forest, savanna/grassland/shrubland, and cropland fires (Table S6). Accounting for the median in-plume ATom HCOOH enhancement of 36 ppt/ppb CO based on these precursor emissions would thus require an average carbon-based yield of 19–40%—much higher than has been reported^{5, 16, 130, 131}. Primary ERs for submicron aerosol OC are estimated at 98–226 ppt C/ppb CO (Table S6); in this case an average per-carbon HCOOH yield of 16–37% from aerosol OC would be needed to explain a HCOOH enhancement of 36 ppt/ppb CO. Malecha and Nizkorodov (2016)²³ report based on chamber measurements that OA mass is photochemically converted to HCOOH at a rate of 1.9/yr. Assuming that photodegradation is the major loss pathway for biomass burning OA (at a similar loss rate as derived in chamber experiments for biogenic/anthropogenic secondary), the 1.9/yr corresponds to a carbon-based HCOOH yield of ~0.3–3% over 1–10 days of aging, which again is far lower (>10–100×) than required to account for the ATom observations. Together, the above calculations indicate that additional, unknown production pathways are needed to account for the HCOOH enhancements observed in many of the aged fired fire plumes sampled by ATom.

4. Summary and implications

The ATom campaign sampled from pole-to-pole over the remote Pacific and Atlantic Oceans. Outside of continental plumes, HCOOH concentrations were generally less than 100 ppt with no consistent latitudinal gradient that was detectable. Observed background concentrations in the free troposphere are consistent with prior aircraft profiling of the remote atmosphere and with GEOS-Chem model predictions. We conclude that the missing HCOOH precursors identified in earlier work must be mainly terrestrial in nature and short-lived so that their effects do not manifest throughout the remote atmosphere. Similarly, the missing acetaldehyde source in the remote background atmosphere inferred by Wang et al. (2019)⁴⁰ apparently does not lead to substantial background HCOOH production.

ATom observations provide signatures of HCOOH removal via wet scavenging and ocean uptake. The contrasting behavior between HCOOH and HCN observed in these situations qualitatively matches expectations in the case of wet scavenging, with stronger removal for the more-soluble HCOOH. The GEOS-Chem model appears to be underestimating HCOOH removal via wet scavenging, and this is consistent with a small but positive model bias in many parts of the clean mid-to-upper troposphere. In the case of ocean-atmosphere exchange, we find that while HCN is consistently depleted in the marine boundary layer, this is only sporadically the case for HCOOH—despite its strong air-to-sea partitioning tendency and seawater undersaturation. We attribute this to that ongoing secondary production in continental outflow that often exceeds the rate of HCOOH dry deposition to the ocean; other processes related to the sea surface^{38, 39} may also be playing a role.

HCOOH enhancements observed during ATom correlate most strongly with BC and are also closely related to the trajectory-derived time since fire influence. These findings reveal a significant role for fires in driving remote HCOOH variability and highlight the general influence of biomass burning in the remote atmosphere⁵⁷. Normalized excess mixing ratios (NEMRs) for HCOOH in the fire plumes sampled during ATom are frequently 1–2 orders of magnitude higher than expected solely from its primary emissions, implying substantial secondary production in fire plumes (as found in prior near-source and satellite-based studies). We find that HCOOH is a major reactive carbon reservoir in aged fire plumes (up to 33% of the total gas+particle-phase reactive organic carbon, ROC). Such enhancements are larger than can be explained based on its known pyrogenic precursors and current understanding of its formation pathways.

The range of HCOOH NEMR values across the 38 mainly-pyrogenic ATom plumes identified here is consistent with findings from prior satellite-based studies, highlighting the potential for using satellite data in future investigations of fire plume aging. Unusually high HCOOH NEMRs observed in several mixed anthropogenic+fire plumes sampled during ATom may reflect the vigorous oxidation that can occur when biomass burning VOCs mix with elevated urban-derived NO_x. Further research is needed to better understand the chemistry and associated air quality impacts during such urban-fire interactions, which can lead to extreme air pollution events for cities in wildfire-prone areas such as the western US.

Fire plumes sampled during ATom range in age up to ~10 days. While the largest HCOOH enhancements occur in the aged plumes, observed HCOOH NEMRs are highly variable and do not correlate consistently with plume age. What drives this variability? Answering this question requires dedicated in-situ plume Lagrangian sampling that is chemically comprehensive and spans a wider age range than typically covered in field studies (<10 hrs). Such extended sampling is especially important for relatively long-lived species such as HCOOH (lifetime of ~weeks). Satellite observations^{132, 133}, while more limited in species coverage, also have potential for sampling across the entire plume evolution spectrum.