On the C130, aerosols were sampled through a low-turbulence, isokinetic inlet and passed to each instrument. Total particle concentration is measured with two condensation particle counters at 1 Hz with lower cutoff sizes of 10 nm (CPC model 3772, TSI Inc., St. Paul, MN) and 3 nm (CPC model 3025, TSI Inc., St. Paul, MN). A laser aerosol spectrometer (LAS model 3340, TSI Inc., St. Paul, MN) measures particle optical diameter distributions between 100 and 3500 nm at 1 Hz. The LAS particle concentration is integrated to identify the number of particles greater than 100 nm. Submicron particles are also analyzed with a high-resolution time-of-flight aerosol mass spectrometer (AMS, Aerodyne Research Inc., Billerica, MA) (DeCarlo et al., 2006) that measures non-refractory inorganic (sulfate, ammonium, nitrate, chloride) and organic components at 30 s intervals. To identify anthropogenic pollution, refractory black carbon particle mass is measured with a single-particle soot photometer at 0.1 Hz (SP2, DMT, Boulder, CO), and carbon monoxide mixing ratio is measured with a CO/CO2 gas analyzer that employs a cavity-enhanced absorption technique at 1 Hz (LGR, San Jose, CA). CPC, LAS, AMS, and SP2 measurements are all reported with respect to standard temperature and pressure (273.15 K, 1013 hPa). A cloud droplet probe (CDP, DMT, Boulder, CO) measured cloud droplet size distributions for droplets ranging from 2 to 50 µm in diameter, and a cloud imaging probe (CIP, DMT, Boulder, CO) measured cloud droplet size distributions for droplets ranging from 50 to 1600 µm in diameter. The nadir-pointing high-spectral-resolution lidar (HSRL) measured the vertical curtain of aerosol backscatter coefficient (532 nm wavelength) as well as cloud-top heights (Hair et al., 2008). The dimethylsulfide (DMS) mixing ratio was measured using a proton-transfer-reaction time-of-flight mass spectrometer (PTR-ToF-MS) at 1 Hz (Müller et al., 2014). Air temperature was measured with a non-deiced total temperature sensor (model 102, Rosemount, St. Louis, MO), and total and static pressure was measured with a flush-mounted static pressure sensor and total pressure sensor (MADT 2014, Rosemount, St. Louis, MO). Wind components were measured with differential pressure measurements across a five-hole pressure port system configured on the C130 nose radome. Wind-component measurements were corrected for aircraft altitude and inertial motion.

R/V Atlantis aerosol instruments sampled air through a temperature-controlled, isokinetic inlet mounted on the forward O2 deck of the ship (∼18 m a.s.l.). Collected air was subsequently dried using silica gel diffusion dryers before the flow was distributed to the instruments. A 1.0 µm sharp cut cyclone (SCC 2.229, BGI Inc. US) removes large, coarse mode particles (mainly sea salt) in order to capture only the submicron aerosol fraction. Particle number concentrations above ∼13 and ∼5 nm were measured using two condensation particle counters (CPC 3010/3785, TSI Inc., St. Paul, MN), while a scanning electrical mobility sizer (SEMS, model 138, 2002, BMI, Hayward, CA) measured dry particle size distributions (0.01–0.9 µm diameter). A differential mobility particle sizer (DMPS, University of Vienna) (Winklmayr et al., 1991) was used to measure the number size distribution of dry submicron (0.02–0.8 µm diameter) ambient particles when SEMS measurements were unavailable. Accumulation mode particle number concentrations (diameters >100 nm) are calculated by integrating the portion of the SEMS and the DMPS size distributions above 100 nm. A single-particle soot photometer (SP2, DMT, Boulder, CO) is used to measure refractory black carbon mass concentration. A second AMS (same model as the one on the C130) is on the R/V Atlantis to measure non-refractory inorganic (sulfate, ammonium, nitrate, chloride) and organic particles (DeCarlo et al., 2006). A dual-flow-loop two-filter radon (222Rn) detector provides information on air mass origin and, specifically, continental influences (Whittlestone and Zahorowski, 1998). DMS is measured with a chemical ionization mass spectrometer (Bell et al., 2013, 2015).

We use the FLEXible PARTicle dispersion model (FLEXPART; Stohl et al., 2005) to assess air mass back trajectories and to identify cases in which we can link upwind particle measurements to subsequent downwind measurements (Zhang et al., 2014). Back trajectories of 10 d (where d stands for day) with 6 h intervals are computed for the R/V Atlantis cruise track every hour and the C130 flight path every 20 min. The Global Forecast System (GFS) and its final analysis (NCEP/NWS/NOAA/USDC, 2000) with 3 h resolution, 1∘ horizontal resolution, and 26 vertical levels are used to drive all simulations. Each simulation consists of 10 000 passive particle tracers released at the R/V Atlantis or C130 location, and the advection and dispersion of the particles are simulated backwards in time. Positions of these tracers are used to compute a gridded distribution of particle residence times (i.e., the average time an air parcel stays within a model grid cell). In our analysis we average the latitude and longitude of the 6 h interval gridded distributions to produce a trajectory line. The vertical structure of the residence time is column-integrated over only the vertical levels that are completely or partially within the MBL based on Global Data Assimilation System (GDAS) MBL heights. Remaining vertical levels were excluded from analysis. More details about the FLEXPART trajectories and the application to study the impact of marine biogenic particle on cloud can be found in Sanchez et al. (2021).

Merged satellite products, at 0.25∘ horizontal resolution (chlorophyll a and cloud fraction), are obtained from the GlobColour project (Maritorena et al., 2010; Maritorena and Siegel, 2005). Here, we use chlorophyll a (chl a) as a simple proxy for phytoplankton biomass (Behrenfeld et al., 2016; Lyngsgaard et al., 2017; Meskhidze and Nenes, 2010; Pastor et al., 2013). GOES-13 visible satellite imagery is used to provide a perspective of cloud coverage relevant for the C130 flights, the R/V Atlantis ship track, and the FLEXPART back trajectories. Sea surface temperatures from level 4 data products, derived from optimally interpolated multi-sensor high-resolution datasets, are available from the Group for High Resolution Sea Surface Temperature (GHRSST) (ABOM, 2008). Finally, 6 h instantaneous horizontal wind vectors at 985 mbar are obtained from the Modern-Era Retrospective analysis for Research and Applications version 2 (MERRA-2) and used to determine seasonal patterns in wind direction (Gelaro et al., 2017; Global Modeling And Assimilation Office, 2015). Satellite-derived surface currents are obtained from Ocean Surface Current Analysis Real-time (OSCAR) (Bonjean and Lagerloef, 2002; ESR, 2007).

The time series of the R/V Atlantis particle and DMS concentrations as well as particle composition from 17 through 19 September 2017 is shown in Fig. 2, where the two highlighted boxes emphasize the transitions in measured aerosol regimes. The first transition is due to the passing of an occluded front with a region of open-cell clouds following behind, as seen by GOES images in Fig. 1a–e. The second transition involves the ship entering a region influenced by continental air, as evident from the elevated radon concentrations (>500 mBq m-3) (Fig. 2d). Between these two transitions the ship remains in the system of open-cell clouds (Figs. 1a–e, S1). In this open-cell region, the particle mass concentration increases as the R/V Atlantis travels southwest.

The aircraft measurements from FLT17 are conducted during 10:25–20:35 UTC on 17 September, upwind of the ship and FLT19 measurements conducted hours to days later, as discussed in detail in Sect. 3.3. The black carbon concentration on both flights is consistent with previously identified clean marine conditions (<25 ng m-3) (Saliba et al., 2020; Sanchez et al., 2021), and similarly the observed carbon monoxide is consistent with observations made at an eastern North Atlantic research facility that is dominated by clean marine conditions (<130 ppb) (Zheng et al., 2018) for near-surface measurements made on both flights (Figs. S2 and S3). Carbon monoxide has a lifetime of about 1 month (Seinfeld and Pandis, 2006) and, therefore, is normally elevated in the North Atlantic relative to polar regions and other remote ocean regions that are not as often directly upwind of continental regions.

Figure 3 shows C130 marine boundary layer particle, cloud droplet number concentration (CDNC), and HSRL backscatter coefficient as a function of longitude for the roughly east–west flight paths shown in Fig. 1c–e. After arriving at the ship, the aircraft initially transited west at low altitude making in situ measurements before spiraling up to high altitude near 57∘ W and retracing the same flight path with the HSRL. Because the aircraft was continually ascending and descending throughout the marine boundary layer during the in situ sampling flight portion, we group the data into statistical boxes that represent the near-surface horizontal clear air legs (∼0.1 km altitude) for the aerosol shown in Fig. 3b and c. Similarly, the cloud measurements (Fig. 3d–f) are grouped for each in-cloud horizontal leg (∼1.0–1.5 km altitude) and also the vertical profile at ∼39.5∘ W. The nature of collecting CDNC along a linear flight path can introduce a bias if measurements are disproportionately collected on cloud edges (downdraft regions) or cloud cores (updraft regions). For this reason, Table 1 includes several summary CDNC and updraft velocity statistics for each cloud leg. The CDNC geometric mean limits the influence of outliers relative to a simple arithmetic mean, while the updraft weighted CDNC removes the inclusion of measurements in downdrafts where CDNC may be decreasing due to evaporation, and, finally, the 90th percentile likely represents an approximate value of the CDNC in an adiabatic updraft core. The weighted updraft is also included to show that the change in CDNC is not simply due to a change in updraft velocity. For example, the weighted updraft velocity at 53.0 and 44.7∘ W is 0.34 and 0.33 m s-1, respectively, but the CDNCs between the two locations vary by a factor of 3–5. Figure 3a shows the C130 altitude from the low-altitude aircraft flight track overlaid on the high-altitude remote HSRL backscatter curtain over the same flight track, which was obtained during a high-altitude, remote sensing leg that was conducted immediately after the near-surface in situ measurement flight legs with minimal delay. It is apparent from Fig. 3e that the westernmost two in-cloud legs (at 53 and 55∘ W) tended to have average CDNCs that were approximately 2–4 times higher than those observed on the eastern part of the track. The CIP measurements in Fig. 3d show that the clouds were precipitating, with peak precipitation sized droplets near the open-to-closed-cell transition, indicating there was removal of aerosol and cloud droplets through precipitation scavenging. This is consistent with previously published observations (Abel et al., 2017; Mechoso et al., 2014; Wood et al., 2011) and simulations (Mechoso et al., 2014; Wang and Feingold, 2009; Yamaguchi et al., 2017), showing reductions in CDNC ranging from 50 % to 90 %. These two regimes are identifiable in both the GOES imagery (Fig. 1), as well as the HSRL curtain (Fig. 3a), where the highest observed CDNCs are in the closed-cell cloud regime to the west, while the lowest CDNCs are in the open-cell cloud regime to the east. Intermediate CDNCs are apparent in the transitional region in between. While the overall aerosol number concentration (N>3nm) does not have a clear trend throughout this period, the accumulation mode aerosol number concentration (N>100nm) (Fig. 3c) and bulk organic and sulfate particle mass concentration (Fig. 3b) are lower on the eastern part of the flight, consistent with the lower CDNCs. There is no clear trend in the observed LWC (Fig. 3f), other than the fact that the highest LWCs were higher in the open-cell region, where cloud tops were higher. The HSRL particle backscatter coefficient measurements show the westernmost near-surface horizontal leg occurred in a cloud-free zone west of the closed-cell cloud regime (Fig. 3a). Notably, the observed near-surface particle concentrations and the back trajectories initialized in the cloud-free and closed-cell regimes are similar (53–59∘ W in Figs. 3b, c, S4), demonstrating that both regimes contain similar particle populations and have a similar source, despite differences in cloud coverage.

Figure 4 shows C130 marine boundary layer aerosol, cloud droplet measurements, and HSRL back scatter coefficient as a function of longitude (similar to the FLT17 case in Fig. 3) for the flight paths shown in Fig. 1g–j. Similar to Table 1, Table 2 includes CDNC and updraft velocity statistics for each cloud leg, showing the trend along the flight path. Note the portion of FLT19 shown in Fig. 1i has a significantly larger northward component compared to the low-altitude portion of the flight path shown in Fig. 1j, causing the higher density of vertical profiles and horizontal leg measurements between 37 and 40∘ W in Fig. 4 and Table 2. Figure 1 satellite images and Fig. 4a HSRL backscatter measurements show that none of the near-surface or in-cloud measurements were in the closed-cell region of the cold-air outbreak. Similar to the FLT17 case, the HSRL curtain covered roughly the same horizontal location as the C130 surface measurements in Fig. 4 (data are not reported here for two periods where the aircraft deviated from the subsequent low-altitude flight track). For FLT19, the high-altitude remote sensing leg was conducted immediately before the in situ surface measurements. A key feature to note in Fig. 4 is the significant change in average particle concentration and CDNC to the west of 40∘ W. Also, the clouds associated with the high particle and CDNC are non-precipitating (Fig. 4d). Further analysis in Sect. 3.3.2 suggests that the higher particle and CDNC region is downwind of the cloud-free region measured on FLT17 and not the cold-air outbreak. In the remaining sections, we explore the cause of the differing adjacent meteorological conditions and their impact on downwind aerosol and cloud properties.

Figure 5a and b present vertical profiles of potential temperature corresponding to the aircraft inline ascents and descents shown in Fig. 3a and Fig. 4a for FLT17 and FLT19, respectively. Only continuous vertical profiles are included in Fig. 5, though some are excluded to prevent substantial overlap of multiple similar vertical profiles. The square points at the sea surface (0 km altitude) represent the satellite-measured SST. The color of each point in Fig. 5 represents the longitude at which the measurement was made. For FLT17, shown in Fig. 5a, the marine boundary layer is neutrally buoyant to the west and becomes increasingly decoupled farther to the east (with the exception of the western most vertical profile at 58∘ W). This evolution in the profile stability and particle concentration is consistent with cold-air outbreaks where processes in the closed-cell region cause the marine boundary layer to decouple, leading to the transition from closed- to open-cell clouds (Abel et al., 2017; Albrecht et al., 1995; Christensen et al., 2020; de Roode et al., 2016; Sandu and Stevens, 2011; Wood et al., 2011; Yamaguchi et al., 2017). None of the vertical profiles from FLT19 are in the closed-cell region, but the profiles to the west, nearest the closed-to-open-cell transition of the cold-air outbreak, are decoupled and stable (Fig. 5b), consistent with observations in FLT17. However, the neutrally buoyant boundary layers to the east of 40∘ W are not consistent with this pattern. These boundary layer profiles are neutrally stable, despite being in an area of cumulus clouds because the eastern portion of the flight is not actually part of the open-cell region of cold-air outbreak or even downwind of it (discussed in Sect. 3.3.2) and is therefore not influenced by meteorological processes associated with the cold-air outbreak.

The FLT17 stable vertical profile at about 58∘ W (Fig. 5a) is from the clear-sky region that is to the west of the closed-cell region (Fig. 1d). This vertical profile is very stable and strongly decoupled. The cause of the differing atmospheric stability in this region is the low SST (Fig. 5a), which cools and stabilizes the marine boundary layer, preventing the vertical transport of water vapor and cloud formation. For almost all the other vertical profiles in Fig. 5, the SST is greater than the atmospheric temperature near the sea surface, consistent with the processes of advection of cold air over warm SST that lead to the stratocumulus-to-cumulus transitions in cold-air outbreaks (Albrecht et al., 1995; de Roode et al., 2016). The neutrally stable vertical profiles with higher particle concentrations and CDNC in FLT19 (Figs. 5b and 4b, c, e) are actually downwind of this same clear-sky region (further discussed in Sect. 3.3.2). The reason for the characteristic differences in the measurements downwind of the clear-sky and cold-air outbreak regions, despite their initial close proximity, is discussed in the following section examining relevant FLEXPART back trajectories.

Averaged satellite SST measurements for September 2017 (corresponding to the NAAMES3 mission) show a local minimum in SST (highlighted by the black box in Fig. 8a) in the cloud-free region to the west of the observed closed-cell region in FLT17 (Fig. 1c). The southerly flow of the Labrador Current bringing cold water from the Arctic Ocean (Fig. S6) causes the observed minimum in SST. Figure 8b shows the spatially collocated minimum in the average cloud fraction (CF), consistent with the visible satellite images in Fig. 1. Figure 8c shows the linear regression between the September 2017 averaged SST and CF from the boxed areas in Fig. 8a and b. Each point in Fig. 8c represents a pixel-by-pixel comparison from the boxed region of Fig. 8a and b. The spatial correlation between the SST and CF suggests that the SST strongly influences the CF in this region during NAAMES3, likely because the air at the surface cools and stabilizes the atmospheric marine boundary layer (consistent with the stable vertical profile and SST at 58∘ W shown in Fig. 5a). A stable marine boundary layer prevents vertical transport of water vapor, thus trapping moisture near the surface, inhibiting cloud formation in the region, and producing the persistent low CF relative to the surrounding area. Our Lagrangian analysis in Sect. 3.3 demonstrates the significant difference in meteorological impacts on cloud properties and particle concentrations downwind of this very stable boundary layer and the adjacent closed-cell region of a cold-air outbreak.

To identify how consistently the SST influences the CF and meteorology in this region, CF and SST are compared for each month using satellite observations from August 2008 to June 2019. The correlation (Pearson's coefficient) between the CF and SST is consistently highest in the late summer, peaking in September, consistent with NAAMES3 analyzed case studies (Fig. 8d). Even the wind direction from MERRA-2, over the same ∼11-year time period, is typically from the northwest (Fig. S7), parallel to the Labrador Current, which is necessary for measurements in the North Atlantic to be downstream of the highlighted area of interest in Fig. 8. Based on these results and the Lagrangian cases shown, this region is an ideal location to study the effect of cloud processing on marine boundary layer particle properties. The proximity of the cloud-free and cold-air outbreak closed-cell region is close and therefore likely often containing inflows of air from similar sources, with similar particle properties. In addition, the difference in meteorological conditions caused by the SST pattern in the region enables one to compare the evolution of the particles initialized in these two different meteorological regimes over a Lagrangian trajectory. The controlling SST forcing on cloud formation in this region of study suggests changes in North Atlantic cold-air outbreaks, due to changes in climate, could have major implications for cloud and particle properties in the North Atlantic. In a previous modeling study Kolstad and Bracegirdle (2008) have shown cold-air outbreaks in the North Atlantic are predicted to weaken in the late 21st century as the difference in the atmospheric temperature and SST is expected to decrease and therefore will likely affect the influence cold-air outbreaks have on the evolution of cloud and aerosol properties.