The ACE campaign
The Antarctic Circumnavigation Expedition (ACE) was a multidisciplinary campaign aboard the R/V Akademik Tryoshnikov, aiming to investigate the biogeochemistry of the Southern Ocean in unprecedented detail, with a wide suite of measurements spanning marine biology, atmospheric chemistry and meteorology, amongst other research areas.13,37,38
ACE was divided into three legs: Leg 1, from Cape Town (20th December 2016) to Hobart (18th January 2017); Leg 2, from Hobart (22nd January 2017) to Punta Arenas (22nd February 2017); Leg 3, from Punta Arenas (26th February 2017) to Cape Town (19th March 2017).
Ambient air was sampled from an inlet mounted on a container located on the second deck of the vessel at a height of approximately 15 m above the ocean surface.37 The inlet consisted of a heated 2-m stainless steel tube (1’’ outer diameter). Mixing ratios of ambient isoprene were monitored using the iDirac, an autonomous portable gas chromatograph coupled with photoionisation detection (GC-PID)39. The instrument was calibrated frequently (every 3–5 hours) using a gas mixture of isoprene in nitrogen prepared in-house and traceable to the National Physical Laboratory (NPL) primary standards 39. Across the ACE campaign, the limit of detection for isoprene was 30 ppt (or pmol mol− 1) with 10% precision. Overall, just under 36% of the isoprene data recorded was above the LOD of the iDirac. In accordance with Hackenberg et al. 11, values below the LOD were set to half the LOD (15 ppt). All analyses of the isoprene data in this work incorporate this adjustment, unless stated otherwise.
A wealth of data was collected during the ACE campaign. The majority of these datasets are available in the ACE online repository 35. The following were used for the analysis of ambient isoprene. Distance to land was calculated using the coordinates of the ship track at a 5 min resolution 40. Ocean water samples were collected throughout the cruise and later analysed for a number of dissolved species, including trace gases (e.g., isoprene) and fluorometric chlorophyll-a 41 as described by Rodríguez-Ros et al.42 Fluorometric chlorophyll-a was corrected according to Galí et al.43 to allow comparison with satellite retrievals.
Wind speed data underwent a thorough correction to account for air flow distortion due to the ship’s structure 44. Further details on the correction can be found in Landwehr et al.45 Trace gas concentrations, including ozone and carbon monoxide (CO) were taken from Schmale et al.46 A number of meteorological parameters such air and sea surface temperature, and boundary layer height were obtained by interpolating the output from the ERA-5 reanalysis 47, available at 1-hour temporal and 0.25°×0.25° spatial resolution, onto the ACE cruise track 48. Five-day air mass back trajectories were calculated using the Lagrangian analysis tool LAGRANTO 49,50.
Satellite retrievals
Chlorophyll-a concentrations in water at a resolution of 0.0416°×0.0416° were obtained from the Level-3 product of the Moderate Resolution Imaging Spectroradiometer (MODIS-Aqua, 51 last accessed 23/03/2022). These were used to determine the concentration of chlorophyll-a along the ACE track (Fig. 1c and Fig S1b). The presence of persistent cloud cover and sea ice gives rise to many gaps in the retrieved ocean chlorophyll-a over the region of interest, especially at high Southern latitudes. This was alleviated by using 8-day composite retrievals as well as by averaging all pixels in a 9×9 array centred on each position along the ACE track. However, if chlorophyll-a concentrations were not available for ≥ 50% of the pixels in each array, those points were excluded from any subsequent analysis.
Sea ice concentration (SIC) fraction from NOAA/NSIDC 52 was used for the analysis of surface type contribution and to differentiate between the marginal ice zone (MIZ) and the more consolidated pack ice.
Isoprene lifetime calculations
The chemical fate of ambient isoprene is primarily to undergo reaction with the hydroxyl radical (OH) and, to a lesser extent, with ozone and the nitrate radical (NO3) 53. Abundances of all three oxidants are needed to establish the atmospheric lifetime of isoprene at a given location. Ozone concentrations were monitored throughout ACE, while those of OH and NO3 were not measured. For the analysis presented here, modelled OH and NO3 abundances were taken from two global models: the UM-UKCA model 54 and the CAMS reanalysis of atmospheric composition 55, which incorporates meteorological variables from the ERA-5 reanalysis. Isoprene lifetime with respect to atmospheric oxidation, τisop, was calculated as:
τisop = 1 / {k(OH + isop)[OH] + k(O3 + isop)[O3 ] + k(NO3 + isop)[NO3]) } (1)
where k indicates the rate coefficient of the reactions of isoprene with OH, ozone and NO3 (in units of cm3 molecule− 1 s− 1) and [OH], [O3] and [NO3] are the number densities of the hydroxyl radical, ozone and the nitrate radical respectively (in units of molecules cm− 3). Rate coefficients were taken from the IUPAC kinetic database 56. Air temperature from on board measurements was used to calculate the temperature dependence of the rate coefficients for each reaction. τisop was used to adjust the length of each back-trajectory, so that effectively the back-trajectory for an air parcel at night-time (long τisop) will stretch further than one at day-time (short τisop). It is worth noting that using a fixed τisop of 2–3 h (typical at mid-latitudes) would limit all trajectories to the immediate vicinity of the ACE track, missing on the influences of different surface types (e.g., marginal ice), as illustrated in Figs. 2 and S4.
Surface type contribution
The points in the back trajectory for each isoprene measurement were assigned to a particular surface type depending on whether the air mass travelled above land, ocean or sea ice. A land mask from the ECMWF ERA-5 reanalysis (0.25°×0.25°) was used to define “land” and “ocean” surface types in the first instance. Grid cells with land mask values greater than 0.5 were attributed to “land”, the rest to “ocean”. For coordinates assigned to the “ocean” surface type, if their sea-ice fraction (SIC, from satellite retrievals) was between 15 and 85%, they were assigned to the “marginal ice zone” (MIZ) surface type; if their SIC was above 85%, they were assigned to the “pack ice” surface type 21. It was also important to differentiate between different types of landmasses. While still identified as “land” using the routine outlined above, landmasses at latitudes higher than 60 °S should not be lumped with those at lower latitudes as they have much lower isoprene emissions (Fig. 4c). Here we separated landmasses at latitudes north of 60 °S (temperate zone) from those below (Antarctic zone). Ice shelves (e.g., the Ross ice shelf) were considered part of Antarctic (> 60°S) landmasses.
ISOREMS fluxes
Monthly isoprene sea-to-air fluxes were calculated for December 2016, January, February and March 2017 using the ISOREMS approach, specifically the parameterisations of Eq. 1–4. 12
To calculate the emissions of isoprene, we first regridded monthly mean chlorophyll-A (CHLA) concentrations from MODIS-Aqua to the resolution used by ERA5. We chose to use hourly wind speed and SST from ERA5 rather than the monthly means of squared wind speed used by Rodríguez-Ros et al. 12 and so applied a linear interpolation of the monthly CHLA fields to generate data on hourly time points. This CHLA data was then used alongside ERA5 hourly SST data to calculate aqueous isoprene concentration, ISO (Eq. 1 in Rodríguez-Ros et al.12, mean value). The ISO field was then used to calculate isoprene emissions, FISO, following Eq. 2 of Rodríguez-Ros et al.12 which also used ERA5 wind speed and SST. The resulting isoprene emissions were then scaled to give the units of kg m− 2 s− 1 and averaged to give monthly mean values as required by UKESM1, and conservatively regridded to UKESM1 N96 resolution. We only considered the region from 90°S to 30°S in this study.
Emissions were applied as a prescribed field (i.e., simulated model meteorology did not affect emissions rates). As is standard in UKESM1, for each grid cell emissions were linearly interpolated between monthly mean values with a fixed period of 5 days before being stepped to the next value. When a diurnal cycle for marine isoprene emissions was not applied, emissions were constant throughout the day. In the simulations where diurnal cycles were applied to marine isoprene emissions, the same stepping approach was applied but emissions varied between 0 at night and a maximum during daylight hours.
For runs where marine isoprene emissions were scaled (e.g., TI_MI_20x), a universal scaling was applied to all grid cells and time points (i.e., marine isoprene emissions in TI_MI_20x were 20x higher at any time point and location than in TI_MI_MEAN).
In the case of TI_MI_MEAN_D_sink and TI_MI_MEAN_D_sink_100x, the emissions of the dummy species acting as an OH sink were equal to and 100 times those of the marine isoprene emissions in TI_MI_MEAN respectively. Like the marine isoprene emissions in TI_MI_MEAN, the dummy emissions had no diurnal scaling. The dummy species reacted with OH, generating only CO2 to prevent any oxidant regeneration. The rate constant was set as half that of OH + isoprene 56 (given the high reactivity of isoprene with OH), with the same temperature dependence:
$$dummy + OH \to C{O}_{2 } ; k=1.35\times 1{0}^{-11}{e}^{ 390/T}$$
2
UKESM model
All model runs were performed using the United Kingdom Earth System Model v1.0 (UKESM1) in atmosphere-only (AMIP) setup, run at a horizontal resolution of 1.25°x1.875 ° with 85 vertical levels up to 85 km 57, and the GLOMAP-mode aerosol scheme, which simulates sulfate, sea salt, black carbon, organic matter, and dust but does not simulate currently nitrate aerosol 58. In this setup, the inert chemical tracer Sec_Org, which condenses irreversibly onto existing aerosol, was produced at a 26% yield solely from reactions of α-pinene and β-pinene with O3, OH, and NO3 with the enhanced yield applied to account for a lack of SOA formation from isoprene or anthropogenic species 58. All runs used the CRI-Strat 2 chemistry scheme 59 which uses the updated isoprene of Jenkin et al.60
Temperature and horizontal wind fields were nudged 61 in all model runs to atmospheric reanalyses from ECMWF 62 to constrain the simulations to consistent meteorology, thus preventing diverging meteorology from adding to the differences resulting from the chemical mechanisms and replicating the atmospheric conditions experienced when the observations were recorded as closely as possible. Nudging only occurred above ∼1200 m in altitude, and thus the majority of the planetary boundary layer was not nudged. The model runs were atmosphere-only runs with prescribed sea surface temperatures (SSTs). CO2 is not emitted but set to a constant field, while methane, CFCs, and N2O are prescribed with constant lower boundary conditions, all at 2014 levels 63.
The emissions used in this study are the same as those developed for the Coupled-Model Intercomparison Project 6 (CMIP6)64. Anthropogenic and biomass burning emissions data for CMIP6 are from the Community Emissions Data System (CEDS), as described by Hoesly et al. 65. 2014 timeslice emissions were used for anthropogenic and biomass burning emissions. Oceanic emissions were from the POET 1990 data set66, and all biogenic emissions, including terrestrial isoprene, were based on 2001–2010 climatologies from Model of Emissions of Gases and Aerosols from Nature under the Monitoring Atmospheric Composition and Climate project (MEGAN-MACC) (MEGAN) version 2.11. All runs used the same anthropogenic, biomass burning and biogenic emissions with the only difference between the addition of emissions of marine isoprene and the dummy species used as an OH sink in certain runs. These two sets of emissions are discussed in more detail in the ISOREMS flux sections.
Model-Observation Comparison along Ship Track
For each hourly observation data point, we extracted the contemporaneous modelled concentration for the grid cell whose centre coordinate displayed the smallest difference in latitude and longitude to the coordinate of the observational data point.
Diel cycle
Local solar times were calculated by adding a factor of 4 minutes per degree longitude (-180° to 180°) to UTC time 67. To ensure the diel cycle was representative of marine sources alone, the isoprene data points included in the calculations were those for which the ocean contribution along the back trajectory was greater than 99% and the terrestrial isoprene from TI_BASE runs was lower than 1 ppt. To reduce the impact of spikes and short term variability, an additional set of diel cycles (Fig S17) was produced in which every isoprene concentration was normalised to the corresponding daily mean concentration following the procedure described in Wohl et al. 10 Night-time was defined as periods when solar irradiance < 1 W m− 2 (see Fig S20).