Glyoxal (CHOCHO), the smallest dicarbonyl, is produced from the photo-oxidation of non-methane volatile organic compounds (NMVOCs), and plays critical and multifaceted roles in tropospheric chemistry . Glyoxal photolyzes to produce HO₂ radicals , thereby propagating atmospheric oxidation and ozone production . Glyoxal can also heterogeneously react on surfaces of aqueous aerosols and cloud droplets to become part of secondary organic aerosol (SOA) . Once in the aqueous phase, glyoxal may react with ammonium, amines, and amino acids to form light-absorbing organic nitrogenous molecules, which radiatively affect climate . In addition, satellite observations of tropospheric glyoxal VCDs serve as a proxy for NMVOC emissions, offering precursor-differentiating information when analyzed in combination with formaldehyde, another widely used proxy .

Despite its importance, the global budget of atmospheric glyoxal remains poorly understood, limiting its utility as a quantitative tracer for NMVOC emissions and hindering assessments of its chemical and climatic impacts. Early model studies by and estimated global atmospheric glyoxal sources of 45 to 56 Tg yr⁻¹, with 53 % to 70 % produced from the oxidation of terrestrial biogenic NMVOCs, predominantly isoprene. A more recent work by estimated a 47 Tg yr⁻¹ global glyoxal source but with smaller contribution from isoprene and larger contributions from monoterpenes and anthropogenic NMVOCs. In contrast, top-down constraints using glyoxal observations from the SCIAMACHY satellite instrument implied a much larger atmospheric glyoxal source of 107 Tg yr⁻¹ from terrestrial NMVOC emissions and photochemical production, with approximately 26 % attributed to unidentified terrestrial biogenic precursors . More recently, used TROPOMI satellite observations to inversely derive an 110 Tg yr⁻¹ global terrestrial source for atmospheric glyoxal. However, over 40 % of this inferred terrestrial source could not be accounted for by known precursors and chemical mechanisms, highlighting the persistent gap in our understanding of atmospheric glyoxal sources.

Over the past decade, laboratory experiments and theoretical studies have advanced our understanding of the photochemistry of key glyoxal precursors, such as isoprene , aromatics , and other hydrocarbons . However, discrepancies persist between modeled and observed glyoxal abundances and budgets . For example, the global mean glyoxal yield from isoprene oxidation simulated with a recently developed chemical mechanism was 2 % , markedly below previous laboratory measurements and mechanistic calculations . These discrepancies highlight critical gaps in our understanding of glyoxal's formation pathways from NMVOC precursors.

Another major uncertainty in the global budget of glyoxal is the undetermined oceanic contribution. A number of observations detected elevated concentrations of glyoxal in sea water samples (1.4 to 4.8 nmol L⁻¹) and in the marine boundary layer (MBL) air (5.6 to 80 ppt) , with tentative correlations with the dissolved organic carbon (DOC) or chlorophyll-a contents of sea water . These correlations suggested a biogenic source of glyoxal or its precursor in the MBL, although the mechanistic pathways have remained unclear . A few glyoxal precursors, such as isoprene, ethylene, and propene have been observed in the remote MBL, but their reported concentrations or fluxes cannot explain the observed glyoxal abundance in the MBL . Early satellite retrievals also indicated glyoxal enhancements in the tropical MBL, but there were concerns that those glyoxal retrievals may be confounded by water absorption artifacts . Recent satellite retrieval studies have mitigated water artifacts by incorporating temperature-dependent cross-sections and optimized spectral fitting windows, yet they still found glyoxal hotspots over biologically active tropical oceans . In addition, aircraft observations found the enhanced glyoxal to be vertically well-mixed throughout the MBL . Collectively, these observations form a line of evidence for substantial glyoxal production in the MBL from yet unidentified marine precursors. The photochemical impact of this potential glyoxal production in the MBL has not been explored.

In this study, we developed an updated global atmospheric glyoxal simulation with the goal of better supporting atmospheric chemistry research, as well as the use of glyoxal for inverting NMVOC emissions and for diagnosing photochemical regime. We incorporated recent experimental, observational, and theoretical knowledge advancements in four aspects: (1) the emissions of glyoxal and its precursors from terrestrial sources, (2) a potential marine photochemical source of glyoxal, (3) updated production pathways and yields of glyoxal from key precursors, and (4) mechanistic representations of heterogeneous uptake on cloud droplets and aqueous aerosols. We evaluated the simulated atmospheric glyoxal abundances against global surface, aircraft, and satellite observations. Finally, we assessed the impacts of the revised atmospheric glyoxal budget on tropospheric oxidation capacity, ozone, and SOA formation globally.

We first evaluated the atmospheric glyoxal simulation in the standard GEOS-Chem (GC-CTRL simulation) against global observations to identify aspects for model improvements. Figure  compares the annual mean (July 2019–June 2020) tropospheric glyoxal VCDs from the GC-CTRL simulation against TROPOMI observations. TROPOMI showed prominent glyoxal enhancements over major biogenic source regions (Southeast U.S., East and South Asia, tropical forests of South America and Africa), seasonal burning areas (boreal forests of North America and Asia, and forests, savannas, and farm and peat lands in South America, Africa and Southeast Asia), and major anthropogenic source regions. These spatial patterns reflected the expected glyoxal production from biogenic isoprene, as well as the emissions of glyoxal and its precursors from biomass burning and anthropogenic emissions .

The GC-CTRL simulation broadly reproduced the TROPOMI-observed spatial patterns over land (Fig. a, b), but the simulated annual mean glyoxal VCDs were systematically lower than TROPOMI observations, with normalized mean biases (NMB) of -77 % globally and -60 % over land (Fig. S2). In particular, the model substantially underestimated glyoxal over the major source regions of biogenic isoprene in the tropics and the mid-latitudes. Figure  evaluated the simulated concentrations of glyoxal against individual ground-based and aircraft measurements over land (Table S1). Some measurements were made in urban areas or areas affected by local biomass burning (Table S1) and could not be reproduced by our simulations at coarse resolution. However, even with these outliers excluded, the model still showed substantial underestimations of glyoxal concentrations compared to measurements at the surface and in the boundary layer (NMB = -90 %), suggesting systematic errors in precursor emissions and/or near-surface chemistry.

In addition, TROPOMI observed glyoxal enhancements over the tropical oceans (Fig. a) that were spatially correlated with the MODIS Aqua retrieved chlorophyll-a contents of surface sea water (Fig. S3, spatial correlation r = 0.5 over the tropical ocean), potentially indicating a marine biogenic source, which was also implied in previous studies . This feature was completely absent in the GC-CTRL simulation (Fig. b) due to the model's lack of marine emissions of glyoxal or its precursors. The NMB of simulated glyoxal concentrations against TROPOMI observations over the global oceans was -88 % (Fig. b). Lengthening the simulated glyoxal production timescale from terrestrial precursors would not reproduce the observed marine pattern, indicating that the glyoxal enhancements over the oceans could not be explained by the long-range transport of terrestrial precursors alone . Figure  shows that the GC-CTRL simulation was also unable to reproduce the magnitudes and variations of individual glyoxal measurements in the MBL (Table S1, NMB = -92 %). Figure S4 compares the vertical profiles of simulated glyoxal concentrations against aircraft observations over the Northeast Pacific, North Atlantic, Tropical Atlantic, and the South Atlantic . These measurements reported glyoxal concentrations exceeding 10 ppt in the MBL, decreasing gradually with altitude. These observed concentrations of glyoxal in the marine atmosphere were not reproduced in the GC-CTRL simulations.

The evaluation of the GC-CTRL simulation above indicated potential model deficiencies in precursor emissions over land and ocean, as well as in the photochemical production of glyoxal from those precursors. We addressed these issues in the following sections and discussed the uncertainties given the current levels of understanding.

We evaluated our revised atmospheric glyoxal simulations against global satellite and surface measurements and found that incorporating updated precursor emissions and chemistry improved the overall model–observation consistency. Figure  compares the annual mean tropospheric glyoxal VCDs at satellite overpass time (13:30 LT) from sensitivity experiments with TROPOMI observations and the standard GEOS-Chem simulation (GC-CTRL). The GC-TM-E experiment updated the terrestrial precursor emissions and included a hypothetical marine boundary layer glyoxal source. This simulation significantly reduced the global bias against TROPOMI observations (Fig. c, NMB =-20 %, r=0.84 globally), primarily because oceanic glyoxal VCDs were constrained to TROPOMI-observed values. In addition, the GC-TM-E simulation enhanced glyoxal levels over major forested and biomass burning regions across North America, Europe, Asia, South America, and Africa, and Southeast Asia, improving consistency against TROPOMI glyoxal observations over land (NMB =-0.47 and r=0.80). The GC-T-EC simulation, which refined both terrestrial precursor emissions and chemistry, showed further improved spatial correlation with TROPOMI over land (Fig. d), (r=0.86 and NMB =-0.47). However, a global bias remained due to the lack of representation of glyoxal over the oceans.

The GC-TM-EC simulation, which incorporated current-best knowledge on terrestrial precursor emission and chemistry, along with TROPOMI-constrained marine glyoxal source, showed the best overall agreement with global TROPOMI observations (Fig. e). The global bias of simulated annual mean glyoxal VCD against TROPOMI observations was -17 %, with a spatial correlation of r = 0.87. Over land, the NMB was -42 %, with a spatial correlation of r = 0.85. The slight improvement in NMB in the GC-TM-EC simulation relative to the GC-T-EC simulaation over land arose from enhanced glyoxal VCDs over the coastal grids due to the added marine glyoxal source. The remaining land bias in the GC-TM-EC simulation reflected persistent uncertainties in precursor emissions from biogenic, anthropogenic , and biomass burning sources , as well as glyoxal yields from these precursors.

Figure a compares simulated glyoxal concentrations against surface measurements over land. Excluding measurements from urban and biomass-burning sites, the GC-TM-EC simulation showed substantially better agreement with observations over land (NMB = -43 %) than the GC-CTRL simulation (NMB = -77 %). We found that this improvement over land arose not only from refined isoprene emissions and its glyoxal production, but also from a better quantification of glyoxal and its precursors from biomass burning. Figure  compares satellite-observed and simulated VCDs of glyoxal and isoprene, along with the observed and simulated glyoxal-isoprene ratios (RGI) over four high-glyoxal regions. Figure S5 compares the simulated surface isoprene concentrations against measurements in key biogenic source regions. Over North America, the CrIS constraints on isoprene emissions corrected the low bias in GC-CTRL-simulated isoprene VCDs, improving the agreement with observed surface isoprene concentrations, while also improving the agreement between the simulated glyoxal VCDs and the TROPOMI observations. Updates in chemical mechanisms further improved the simulated glyoxal abundances and the RGI, indicating a better representation of glyoxal production from isoprene. In contrast, over South America, Africa, and Southeast Asia, CrIS constraints either reduced or only slightly increased isoprene emissions, which brought the simulated surface isoprene concentrations to better agree with measurements (Fig. S5). However, the enhanced biomass-burning emissions of glyoxal and its precursors substantially raised both the simulated glyoxal abundances and the simulated R_GI to become more consistent with observations. Chemical mechanism updates in GC-TM-EC only brought minor adjustments to simulated glyoxal abundance over these regions. These results underscore the large contribution of biomass burning emissions to atmospheric glyoxal, which must be better constrained to enable accurate NMVOC emission inversions using glyoxal VCDs.

Over the ocean, inclusion of the hypothetical marine glyoxal source markedly improved the agreement of the GC-TM-EC simulation with independent MBL observations. Compared to the GC-CTRL simulation, the normalized mean bias (NMB) against MBL measurements in the tropical Pacific, Southern Ocean, Caribbean Sea, and North Atlantic was reduced from -92 % to 12 % (Fig. b; Table S1). This consistency with observations independent of the TROPOMI retrieval supports the existence of a significant, previously unaccounted–for glyoxal source in the MBL. Figure S4 further compares the simulated glyoxal over the remote ocean with the aircraft observations from and . With the added marine glyoxal source, GC-TM-EC–simulated glyoxal concentrations aligned better with the limited aircraft data. However, the model overestimated glyoxal in surface air over the Northeast Pacific but underestimated it over the Atlantic. Additionally, the aircraft profiles suggested a more gradual decline in glyoxal concentration with altitude than simulated, which may indicate a longer–lived marine–derived precursor. This inference is inconsistent with the current understanding that marine–emitted VOCs are predominantly short–lived alkenes (e.g., ethylene and propene) and isoprene .

Table  summarizes the simulated global burdens and budgets of atmospheric glyoxal from our sensitivity experiments and compares them with the GC-CTRL simulation and . Figure  shows the global patterns of sources and sinks of atmospheric glyoxal in the GC-TM-EC simulation. Assuming that glyoxal and its precursors originated solely from terrestrial sources and using revised chemistry (the GC-T-EC simulation), the estimated global atmospheric glyoxal burden was 15 Gg, 50 % higher than the standard GEOS‐Chem value (10 Gg). This revised burden coincided with the 15 Gg reported by , despite large differences in model emissions and chemical mechanisms. The global atmospheric glyoxal source in the GC-T-EC simulation was 40 Tg yr⁻¹, a 74 % increase over the 23 Tg yr⁻¹ in the GC-CTRL simulation. This enhancement resulted from both elevated primary glyoxal emissions from biomass and biofuel burning, as well asstrengthened secondary chemical productions (67 % enhancement), particularly from isoprene.

Our revised global budget still pointed to isoprene as the most important terrestrial precursor of atmospheric glyoxal, contributing 14 Tg yr⁻¹ (40 %) to the total atmospheric glyoxal source, 79 % stronger than the 7.8 Tg yr⁻¹ in the GC-CTRL simulation. In the GC-T-EC simulation, 86 % of the global glyoxal production from isoprene was through glycolaldehyde, whose global effective yield of glyoxal was 10 %, twice of that in the GC-CTRL simulation. Our revision of the isoprene chemistry increased the global glyoxal yield from isoprene to 3.2 % in the GC-T-EC simulation, compared to the 1.7 % in the GC-CTRL simulation. This yield of glyoxal from isoprene in the global atmosphere was smaller than that predicted by box models, because the key intermediates (such as MVK, IEPOXs, and HPALDs) were removed by deposition or aerosol uptake, limiting their gas-phase reactions . The total glyoxal production from biogenic precursors was 18 Tg yr⁻¹, 45 % of the total atmospheric glyoxal source. The anthropogenic and biomass burning precursors, including aromatics, acetylene, and ethylene, collectively contributed 41 % of the total glyoxal source.

Atmospheric glyoxal was removed predominantly by photolysis (58 %), followed by OH-oxidation (16 %), irreversible uptake by aqueous particles (14 %), and deposition (14 %) (Fig. ). In addition, the heterogeneous removal of glyoxal by aqueous particles was 5.6 Tg yr⁻¹, an order of magnitude larger than that in the GC-CTRL simulation. This increase was due to a combination of increased global glyoxal formation and enhanced irreversible uptake coefficient of glyoxal over land. The simulated global lifetime of glyoxal in the GC-T-EC experiment was 3.4 h, shortened compared to the 3.8 h in the standard GEOS-Chem due to accelerated removal by OH and heterogeneous uptake.

The inclusion of a hypothetical marine glyoxal source in the GC-TM-EC simulation substantially altered the global glyoxal budget, elevating the burden to 39 Gg and the total source to 106 Tg yr⁻¹. Approximately 29 Gg of this enhanced burden was over the ocean, representing a 5.5-fold increase compared to the simulated glyoxal burden over the ocean in the GC-T-EC simulation. Sixty percent (66 Tg yr⁻¹) of the total glyoxal source in the GC-TM-EC simulation was produced in the MBL (Table , Fig. ), constituting a large source of OVOCs in the MBL that could influence the oxidative capacity of the marine atmosphere. Terrestrial source contributions remained similar to those in the GC-T-EC simulation.

The relative strengths of glyoxal sinks also shifted significantly in the GC-TM-EC experiment. Photolysis became an even more important sink for global glyoxal due to the high solar radiation and glyoxal concentrations in the tropical MBL. In contrast, glyoxal removal by heterogeneous uptake became relatively less important, as the added glyoxal was concentrated in high-humidity MBL where the reactive uptake coefficient of glyoxal (γG) was suppressed . Overall, the simulated lifetime of atmospheric glyoxal in the GC-TM-EC experiment was 3.2 h, similar to that in the GC-T-EC experiment.

Finally, we examined how the revised atmospheric glyoxal simulations affected global atmospheric chemistry. Figure  compared the simulated global surface concentrations of OH, HO₂, and O₃, along with the surface OH reactivity (OHR) associated with NMVOCs, in the model experiments. In the GC-T-EC experiment, regional changes to biogenic and biomass burning emissions and modifications to precursor chemistry strongly affected atmospheric chemistry over land. The increased regional isoprene and biomass burning NMVOC emissions in the GC-T-EC simulation strengthened OH reactivity across North America and Euroasia, while reduced regional isoprene emissions decreased OH reactivity over South America and Australia (Fig. h). The simulated surface OH decreased or increased in accordance to these regional changes in OH reactivity, averaging a -4.2 % reduction in global surface OH relative to the GC-CTRL experiment (Fig. h). These changes in OH-initiated oxidation of NMVOCs subsequently altered the photochemistry that produced HO₂, particularly over relatively high-NO_x areas: surface HO₂ concentrations increased by more than 10 % over Europe, the Middle East, and East and South Asia, while decreasing slightly over the Amazon and Australia. Globally, the updates of precursor emissions and glyoxal chemistry increased the global surface O₃ concentration by 17 %.

The revised precursor emissions and glyoxal chemistry also affected global SOA simulation. Simulated total SOA formation increased from 115 Tg yr⁻¹ in the GC-CTRL experiment to 133 Tg yr⁻¹ in the GC-T-EC experiment and 136 Tg yr⁻¹ in the GC-TM-EC experiment. Glyoxal-derived SOA rose from 0.6 Tg yr⁻¹ in the GC-CTRL experiment to 6.4 Tg yr⁻¹ in the GC-T-EC experiment and 7.6 Tg yr⁻¹ in the GC-TM-EC experiment, respectively. Including the hypothetical marine glyoxal source contributed an extra 1.2 Tg yr⁻¹ of glyoxal SOA. Notably, the relative increase in SOA after inclusion of the hypothetical marine glyoxal source was smaller than the corresponding relative increase in gas-phase glyoxal. This was because glyoxal SOA formation is limited by heterogeneous uptake, which was suppressed in high-RH regions by the salting-in effect, thereby reducing SOA production efficiency over humid marine environments .

The hypothetical MBL source of glyoxal also affected the atmospheric oxidative capacity over the ocean, especially in the tropics (Fig. c, f, i, l). The additional glyoxal removed OH in the MBL, while its photolysis produced HO₂. Relative to the GC-T-EC experiment, the added marine glyoxal source (simulated in GC-TM-EC) had a negative impact on the surface OH concentration globally (-0.6 %), especially over the tropical ocean (-1.0 %). The marine glyoxal increased the surface concentration of HO₂ by 6.5 %, the surface NMVOC OH reactivity by 1.9 %, and the surface ozone concentration by 0.3 %, over the tropical ocean.

Figure S6 compares our simulated HO_x vertical profile against the observations in the marine boundary layer (below 2500 m) during the ATom-2 mission . The OH concentrations simulated in the GC-CTRL simulation agreed well with observations, and our improved glyoxal simulation did not significantly change the OH profile over the remote ocean. The HO₂ concentration simulated in GC-TM-EC slightly alleviated the underestimation in GC-TM-E by 5 % but still underestimated HO₂ compared to the observations, indicating the potential lack or persistent underestimation of HO₂ sources (e.g., OVOCs) over the remote MBL.

We also evaluated the broader influence of the revised glyoxal and precursors representations on tropospheric chemistry (Figs. S7, S8, S9). Relative to the GC-CTRL simulation, increased precursor emissions in the GC-TM-EC simulation substantially raised the global mean surface CO concentrations from 99.1 to 113.5 ppb, reducing the model bias against MOPITT observations from -25 % to 13 % (Fig. S7), especially over Eastern United States, East and South Asia, and the tropical South America and Africa. Similarly, enhanced terrestrial precursor emissions moderately increased the simulated global mean surface ozone concentration over land from 22.8 ppb in GC-CTRL to 29.5 ppb in GC-TM-EC, closer to the 34.1 ppb reported in the data–assimilated product of (Fig. S8). The increased precursor emissions and atmospheric glyoxal abundance in the GC-T-EC experiment also led to a stronger heterogeneous uptake of glyoxal by aqueous particles, thereby enhancing SOA formation (Table ). Over China, for example, the GC-T-EC experiment simulated significantly higher SOA concentrations than the GC-CTRL experiment, especially at suburban and remote sites (Fig. S9). These increases brought simulations closer to – though still substantially below – observed concentrations, indicating that SOA formation involves complex mechanisms beyond glyoxal chemistry . The apparent improvements in simulated CO, ozone, and SOA should therefore be interpreted as supporting the chemical plausibility of the revised glyoxal budget and underscoring glyoxal's role as a proxy for atmospheric NMVOC emissions.

We developed an improved global simulation of atmospheric glyoxal using the GEOS-Chem model, advanced by updated emissions and chemical mechanism and consistent with global observations. Assuming glyoxal precursors were entirely terrestrial, we estimated a global atmospheric glyoxal source of 40 Tg yr⁻¹ and a global burden of 15 Gg, effectively improving the consistency between simulated glyoxal VCDs and TROPOMI satellite observations. Isoprene was identified as the dominant terrestrial precursor, while anthropogenic and biomass burning sources – including acetylene, ethylene, and aromatics – collectively contributed 39 % of the global glyoxal source. These results highlighted the importance of non-biogenic precursors to ambient glyoxal in urban and biomass burning areas, with implications for the use of glyoxal for NMVOC emission inversion in such environments.

Even with our improved precursor emission estimates and chemical representation, our simulated glyoxal remained approximately 40 % lower than satellite glyoxal retrievals and surface observations over land. This discrepancy likely stems from both satellite retrieval uncertainties and persistent uncertainties in emissions and chemistry. Anthropogenic emissions of glyoxal precursors, particularly aromatic compounds, are variable and uncertain among different emission inventories and demonstrated a consistent low-bias compared to observations . In addition, while we scaled isoprene and biomass burning emissions by region on an annual basis, this approach may introduce biases across different land cover types and seasons . Uncertainties in simulated OH and NO_x concentrations may moderately affect glyoxal production, but they have a potentially large impact on top-down isoprene estimates, particularly over the Amazonian rainforest, where underestimated OH and NO_x lead to isoprene overestimation for a given emission rate . Further uncertainties arise from incomplete representation of key chemical pathways, including glyoxal production from isoprene under low-NO_x conditions , ozonolysis via Criegee intermediates , detailed isoprene nitrate chemistry , and heterogeneous uptake processes of glyoxal .

Our study highlights the need for improved measurement of VOCs in the MBL and for a deeper exploration of their photochemical transformations. Our exploratory addition of a hypothetical secondary glyoxal source (66 Tg yr⁻¹) raised the global atmospheric glyoxal source to 106 Tg yr⁻¹ and its burden to 39 Gg. While this addition shows tentative consistency with in situ MBL observations, this hypothetical glyoxal source cannot be explained by known marine precursor emissions. Recent work suggests marine biogenic emissions of precursors, such as isoprene, may be larger than previous estimates . However, our evaluation of model results against limited aircraft measurements points to the possible influence of a more long-live precursor, the nature and impacts of which remain highly uncertain. Resolving this gap in marine photochemistry is essential for quantifying the roles of glyoxal and its precursors in marine and global atmospheric chemistry.