Atmospheric isoprene measurements reveal larger-than-expected Southern Ocean emissions

The study investigates the sources and magnitude of marine isoprene emissions over the Southern Ocean (SO) and their atmospheric implications. Isoprene (C5H8) is a highly reactive volatile organic compound (VOC) that influences atmospheric oxidative capacity (notably via OH) and climate (through secondary organic aerosol and cloud processes). While terrestrial emissions dominate globally, marine isoprene emissions are uncertain due to limited in situ data and incomplete understanding of oceanic sources and sinks, including biological production, consumption, and possible abiotic photochemical production in the surface microlayer (SML). Prior estimates of marine isoprene emissions vary by two orders of magnitude, and atmospheric observations in the SO are sparse. The authors analyze shipborne atmospheric isoprene observations from the Antarctic Circumnavigation Expedition (ACE, austral summer 2016–2017), assess source regions (including the marginal ice zone, MIZ), and use the UK Earth System Model (UKESM1) with various marine emission scenarios to reconcile observations and evaluate impacts on OH. The core hypothesis is that current bottom-up marine isoprene fluxes are too low and that an additional, likely daytime, source (e.g., SML photochemistry) is required to explain observed concentrations and diel patterns over the SO.

• Marine isoprene is primarily associated with phytoplankton production modulated by environmental drivers (temperature, radiation), but can also arise from abiotic photochemistry in the SML at the sea–air interface. Recent work indicates significant chemical and biological consumption within the surface ocean, potentially comparable to air–sea exchange, complicating emission estimates.
• Bottom-up and top-down marine isoprene emission estimates diverge by ~2 orders of magnitude (~300 GgC/yr to ~11 TgC/yr), reflecting uncertainties in oceanic cycling, flux parameterizations, and sparse measurements.
• Satellite proxies (e.g., chlorophyll-a) have been used to infer dissolved isoprene and emissions (e.g., ISOREMS parameterization) but validation against atmospheric observations remains limited.
• Prior SO observations generally report low atmospheric mixing ratios (<50 ppt) away from land and blooms, with episodic higher values near coasts and biologically active regions.
• Evidence from Arctic/Antarctic MIZ suggests enhanced biogenic VOCs associated with phytoplankton blooms following sea-ice melt, supporting the MIZ as a hotspot.
• Emerging studies report higher-than-expected marine VOC emissions (e.g., benzenoids, butenes) and SML photochemical sources, which may help explain observed VOC oxidation products (glyoxal, methylglyoxal), organic aerosol contributions, and OH budget closure over marine regions.

Field observations (ACE campaign):

• Platform and period: Antarctic Circumnavigation Expedition aboard R/V Akademik Tryoshnikov during austral summer 2016–2017, divided into three legs (Leg 1: 20 Dec 2016–18 Jan 2017; Leg 2: 22 Jan–22 Feb 2017; Leg 3: 26 Feb–19 Mar 2017).
• Air sampling: Inlet on second deck (~15 m a.s.l.), 2 m stainless steel tube (1" OD) heated to 20 °C; main inlet flow 13 L/min (residence time <3 s).
• Isoprene measurements: iDirac autonomous GC-PID, sampling via ~50 cm SilcoNert2000-treated stainless tube (1/16" OD, 0.04" ID), 20 mL/min (residence time 1–2 s). Frequent calibration every 3–5 h with NPL-traceable standards. Campaign LOD 30 ppt with 10% precision; values below LOD set to half LOD (15 ppt) following established practice; ~36% of data above LOD.
• Co-measurements and ancillary data: Ship-corrected wind speeds; trace gases (e.g., O3, CO); ERA5-interpolated meteorology (air/sea surface temperature, boundary-layer height); underway seawater dissolved isoprene and fluorometric chlorophyll-a (corrected for comparison with satellite).

Satellite and reanalysis products:

• MODIS-Aqua chlorophyll-a Level-3 at 0.0416° resolution; 8-day composites; 9×9 pixel averaging along track; data gaps removed if <50% coverage.
• Sea ice concentration (NOAA/NSIDC) to define sea-ice regimes and MIZ.

Back-trajectories and source-type attribution:

• Five-day back-trajectories computed with LAGRANTO, adjusted by isoprene lifetime (τisop) to reflect chemical loss: τisop = 1/[kOH[OH] + kO3[O3] + kNO3[NO3]] with rate coefficients from IUPAC kinetics; [OH], [NO3] from UM-UKCA and CAMS, [O3] observed; temperature dependence from onboard measurements.
• Surface-type assignment along trajectories: ERA5 land mask (>0.5 = land), otherwise ocean. For ocean points: MIZ if sea ice fraction 15–85%, pack ice if >85%. Land further split into temperate (north of 60°S) vs Antarctic (south of 60°S including ice shelves).

Emission estimation (ISOREMS) and scenarios:

• Monthly sea-to-air marine isoprene fluxes for Dec 2016–Mar 2017 computed via ISOREMS using MODIS chlorophyll-a and ERA5 SST/winds. CHLA regridded and linearly interpolated to hourly to combine with hourly ERA5 SST and wind in ISOREMS equations; fluxes scaled to kg m−2 s−1, monthly-averaged, and regridded to UKESM1 N96. Region considered: 90°S–30°S.
• Emission variants: baseline terrestrial only (MEGAN-MACC climatology); terrestrial + ISOREMS marine (TI_MI_MEAN); marine scaled ×20 (TI_MI_20x). Diurnal scaling runs applied a daytime-peaked cycle to marine emissions (TI_MI_MEAN_D, TI_MI_20x_D). Additional OH sink sensitivity runs emitted a dummy species equal to (×1) and 100× marine isoprene emissions, reacting with OH at half the isoprene rate and yielding CO2 only (k = 1.35×10−11 exp(390/T) cm3 molecule−1 s−1), with no diurnal scaling.

Modeling (UKESM1):

• Configuration: Atmosphere-only (AMIP), 1.25°×1.875° horizontal resolution, 85 vertical levels (to 85 km). GLOMAP-mode aerosol. CRI-Strat 2 chemistry with updated isoprene mechanism.
• Forcing and emissions: Anthropogenic and biomass burning (CEDS, 2014 timeslice); oceanic (POET 1990); biogenic terrestrial (MEGAN-MACC 2001–2010 climatology). CO2 fixed, CH4/CFCs/N2O prescribed (2014). Marine isoprene emissions prescribed per scenarios above.
• Meteorology: Nudged temperature and horizontal winds above ~1200 m to ECMWF reanalyses; PBL largely free-running. Prescribed SSTs.

Model–observation comparison:

• Extracted hourly modeled concentrations at nearest grid cell to observation points along ship track. Verified modeled horizontal wind fields against observations. Evaluated scenarios via model/observation ratios and spatial–temporal comparisons.

Diel cycle analysis:

• Local solar time computed; retained "marine-only" data (back-trajectory ocean contribution >99% and modeled terrestrial isoprene <1 ppt). Produced diel cycles including normalized variants to reduce spikes. Compared observed diel cycle to model runs without and with diurnal emission scaling.

• Observed atmospheric isoprene over the SO during ACE had mean mixing ratios of ~42 ppt for the whole campaign (Leg 1: 33 ppt; Leg 2: 47 ppt; Leg 3: 38 ppt). Extremely high values up to ~1200 ppt occurred at high southern latitudes during Leg 2 (Ross Sea, 3–9 Feb 2017), with a baseline shift from <LOD (30 ppt) to ~70 ppt.
• Back-trajectory and chlorophyll-a analyses indicate the marginal ice zone (MIZ) in the Ross and Amundsen Seas as a significant source region. Air masses from the MIZ exhibited higher isoprene than those from open ocean and were second only to temperate land influences.
• Terrestrial emissions alone cannot explain observed marine isoprene: simulations with terrestrial-only emissions (including a 2× enhancement south of 30°S) produced negligible isoprene away from land (model/obs ratio <1e−9).
• Adding marine emissions via ISOREMS (TI_MI_MEAN) increased modeled isoprene but remained strongly low-biased (median model/obs ratio ~0.025 along the track). Scaling marine emissions by 20 (TI_MI_20x) substantially improved agreement but still underpredicted on average (median model/obs ratio ~0.49).
• Applying a diurnal cycle to marine emissions had little effect on overall bias but is necessary to reproduce the observed diel pattern: observations in marine-only air masses show higher daytime and lower nighttime isoprene, requiring a daytime-peaked source.
• Sensitivity tests indicate that OH suppression by co-emitted VOCs is insufficient to explain the bias: an extreme dummy OH sink (100× marine isoprene emissions) reduced OH by 9–15% over much of the SO (>15% off South America) and increased modeled isoprene by only ~16.7% (~0.9 ppt), far smaller than the model–observation discrepancy.
• Tests compressing modeled boundary layer isoprene into the lowest model level (to mimic reduced vertical mixing) still yielded low biases (median model/obs ratios ~0.15 for 5-level and ~0.31 for 11-level compression), pointing to emissions as the dominant source of bias.
• The inferred need for at least a 20-fold increase in marine isoprene emissions implies significant regional chemical impacts: with 20× marine emissions, DJFM mean surface OH decreased by ~2.1% across the SO and by >8% regionally (e.g., off eastern South America and phytoplankton bloom regions). An annual simulation with 20× emissions gave a ~1.7% annual mean OH decrease with similar spatial patterns.
• The observed diel cycle and MIZ association support a combined source: biotic emissions from bulk seawater plus a light-dependent SML photochemical source. Reported laboratory SML isoprene fluxes (0.5–20×10−12 kg m−2 s−1) are of the right order to reconcile models with observations.

The findings demonstrate that current bottom-up marine isoprene flux estimates substantially underpredict atmospheric concentrations over the Southern Ocean. After systematically excluding alternative explanations—long-range terrestrial transport, vertical mixing biases, and oxidant (OH) suppression by co-emitted VOCs—the most plausible explanation is missing marine emissions. The diel signature of higher daytime isoprene in marine-only air masses requires a daytime-peaked source, inconsistent with a purely constant flux from mixed-layer biotic production given daytime OH loss. This points to photochemical production at the surface microlayer (SML), likely enhanced near the MIZ where sea-ice melt stabilizes the surface layer and supplies precursors, and where winds can be lower, allowing SML persistence. A combined mechanism of bulk biotic emissions (requiring roughly 20× higher fluxes than ISOREMS) plus light-dependent SML emissions best explains both magnitudes and diel patterns. The chemical implications are notable in this pristine environment: increased isoprene reduces OH by several percent regionally, affecting oxidation capacity and the formation of secondary organic aerosol and cloud condensation nuclei. Given the SO’s role as a proxy for pre-industrial conditions, these findings influence our understanding of baseline atmospheric composition and thus modeled radiative forcing changes from pre-industrial to present-day.

This study provides long-duration atmospheric isoprene observations over the Southern Ocean and shows that existing marine emission parameterizations substantially underestimate atmospheric isoprene, requiring at least a 20-fold increase to reconcile with observations. The diel cycle analysis strongly supports an additional daytime source, consistent with photochemical production in the surface microlayer, especially prominent near the marginal ice zone. Enhanced marine isoprene emissions meaningfully reduce regional OH, with implications for oxidation capacity, aerosol formation, and cloud microphysics in remote atmospheres and for estimates of pre-industrial baselines. Future work should prioritize direct sea–air flux measurements of isoprene across seasons, coordinated measurements within the SML and bulk seawater, broader spatial and temporal atmospheric observations, and improved model representations, including explicit coupling of isoprene to organic aerosol schemes, to constrain sources and atmospheric impacts year-round.

• Seasonal and temporal coverage: Observations and emission constraints focus on DJFM (austral summer); the applicability of a uniform 20× scaling to other seasons is uncertain, though an annual sensitivity run suggests comparable relative OH impacts.
• Detection limits and data coverage: Only ~36% of atmospheric isoprene data were above LOD; satellite chlorophyll-a suffers gaps from clouds and sea ice, particularly at high latitudes.
• Vertical transport evaluation: Lack of updraught/vertical velocity observations prevents robust assessment of modeled vertical mixing; compression tests suggest emissions dominate the bias but do not fully rule out transport contributions.
• Oxidant fields: OH and NO3 were not directly measured; lifetimes and some analyses rely on modeled oxidant fields, though comparisons with Antarctic coastal OH observations suggest reasonable realism.
• Emission parameterization and representativeness: ISOREMS depends on proxies (chlorophyll-a, SST) and may not capture SML photochemical sources; direct flux measurements are lacking. The inferred factor-of-20 scaling is a bulk adjustment with uncertain spatial/temporal variability.
• Model structural choices: Prescribed marine emissions (no meteorology-dependent variability beyond monthly means unless diurnal scaling applied); lack of explicit isoprene SOA formation in the aerosol scheme could under-represent aerosol impacts; PBL not nudged may introduce local meteorological discrepancies.