Sea-salt particles are the largest aerosol component in the atmosphere by mass (Liao et al., 2006). They represent about 30 % of global column optical depth due to aerosols (Bellouin et al., 2013), a somewhat smaller percentage than for mass because of their relatively large size compared to other aerosols. Global climate models disagree on future changes in sea-salt aerosol due to changes in wind speed and sea ice in a warming climate (Liao et al., 2006; Jones et al., 2007; Boucher et al., 2013). Given the large contribution of sea salt to the global aerosol optical depth, this could represent a significant climate feedback, with uncertainty even between different scenarios in the same model (Hoose et al., 2008). Sulfuric acid, nitric acid, and some other acids can displace halogens in salt particles. This provides both a sink for sulfate and nitrate and a source of reactive chlorine, bromine, and iodine to the atmosphere (Chameides and Stelson, 1992; Finlayson-Pitts and Hemminger, 2000).

There is a large literature on the source of sea-salt aerosol as a function of wind speed (e.g., Gong, 2003; Lewis and Schwartz, 2004; Grythe et al., 2014). There has been increased recognition of the importance of submicron salt particles to aerosol number (Kreidenweis et al., 1998; Clarke et al., 2003, 2006). These submicron sea-salt particles are enriched in organics compared to sea water, although the amount of enrichment is not consistent and may vary under differing conditions (Middlebrook et al., 1998; Modini et al., 2010; Vignati et al., 2010; Ovadnevaite et al., 2011; Gantt and Meskhidze, 2013).

Almost all of the sea-salt aerosol literature considers measurements within the marine boundary layer and even there consists mostly of surface measurements. There have been few measurements of how sea salt varies with altitude. Shinozuka et al. (2004) presented profiles of nonvolatile aerosol, presumed to be sea salt, up to about 2 km altitude for one region in the Southern Ocean and one region in the tropical Pacific.

We present here the first measurements of the concentration of sea-salt aerosol over a wide range of altitudes and latitudes. We consider the sea-salt vertical transport, wet removal, and compositional variability. These data provide strong constraints on aerosol transport efficiency out of the marine boundary layer and are a useful tool in evaluating aerosol removal in large-scale models.

We quantify sea-salt aerosol by merging measured size distributions with the fraction of particles in each size range identified as sea salt by single particle mass spectrometry. Sea-salt particles were identified from mass spectra of single aerosol particles using the Particle Analysis by Laser Mass Spectrometry (PALMS) instrument (Thomson et al., 2000). Particles enter a vacuum and cross a split continuous laser beam. The transit time between the beams provides the particle velocity, used to determine its aerodynamic diameter. The aerosol inlet to PALMS is controlled to about 35 mbar, with a small dependence on outside pressure because the pressure transducer used for control was not positioned to capture the full effect of a jet downstream of the first critical orifice. Transit times were calibrated to known particle sizes at laboratory pressure (about 820 mbar) before and after every field deployment. An excimer laser is triggered when a particle arrives at the second laser beam and ions are produced when the 193 nm pulse hits the particle. Either positive or negative ions are analyzed with a time-of-flight mass spectrometer, with the polarity switched every few minutes. For these data, about 60 % of the time was spent acquiring positive ion spectra.

The PALMS instrument was flown on the NASA DC-8 during the Atmospheric Tomography Mission (ATom) as well as earlier missions. The ATom mission consisted of several series of flights between about 85∘ N and 65–80∘ S latitude over both the Pacific and Atlantic oceans. Flights consisted of successive en route ascents and descents from about 160 m to 12 km with 5 to 15 min at the bottom and top of each profile. An extensive payload characterized both gas-phase and aerosol species (ATom, 2017; https://daac.ornl.gov/ATOM/guides/ATom_merge.html, last access: 25 September 2018). Two deployments are considered here: ATom1 from 29 July to 22 August 2016, in northern summer, and ATom2 26 January to 21 February 2017, in northern winter. Some preliminary data from one ATom3 flight (14 October 2017) are included to extend the latitude range further south. During ATom1 and ATom2, PALMS acquired about 510 000 positive ion spectra and 350 000 negative ion spectra of individual particles.

Particle size distributions were measured by an ultra-high-sensitivity aerosol spectrometer (UHSAS; Kupc et al., 2018) and a laser aerosol spectrometer (LAS, TSI Incorporated). In the merged size distribution, UHSAS data were used for particles less than 0.51 µm diameter during ATom1 and 0.97 µm during ATom2; the LAS was used for larger particles. The two instruments agreed extremely well in the overlap region (about 0.4 to 0.97 µm) for all of ATom1 and at low altitudes during ATom2. A leak in the sheath flow of the LAS, traced to the threads of the set screws that center the inlet inside the sheath flow, led to the LAS under-sampling particles during ATom2 at high altitudes, when the aircraft cabin pressure was much higher than the sample pressure. The extra sheath flow reduced the sampling rate but did not introduce false counts, as checked occasionally in-flight with filtered air. A correction was derived by comparing the LAS to the UHSAS, a printed optical particle spectrometer (POPS), and cloud aerosol spectrometer (CAS) data.

The PALMS instrument has substantial biases in sampling efficiency for different size particles, with higher efficiency for particles about 1 to 3 µm diameter compared to about 300 nm by a factor of about 10 and steeply falling efficiency for sizes smaller than 300 nm. In addition, a custom virtual impactor in the PALMS sampling line deliberately enriched the concentration of super-micron particles in order to get better statistics for dust and sea salt. To some extent the bias toward sampling larger particles is beneficial because it enhances the rate of sampling particles with more mass, but it complicates interpretation. Overall, PALMS is much better at measuring fractional rather than absolute abundances of species such as sea salt. Combining such fractions with independently measured size distributions allows quantitative concentration measurements. Rather than directly calibrating the PALMS and virtual impactor sampling efficiencies, data here are normalized to the UHSAS and LAS particle size distributions. Conceptually, if a given percentage of, for example, 1 to 2 µm diameter particles are sea salt, then the absolute concentration can be determined by multiplying the concentration of particles of that size from the LAS by the percentage of sea salt determined by PALMS. Although simple in concept, this normalization is complicated in detail because the UHSAS and LAS measure optical diameter, which depends on refractive index, whereas PALMS measures aerodynamic diameter, which depends on density. The PALMS vacuum inlet is pressure-controlled so its transmission is mostly independent of aircraft altitude. A separate publication (Froyd et al., 2019) is in preparation with more details about the normalization.

PALMS, the LAS, and the UHSAS all sampled from a University of Hawaii inlet owned by NASA Langley on the DC-8 (McNaughton et al., 2007). About 1.5 m of 0.25 in. outside diameter stainless steel tubing with a usual volumetric flow rate of 3.5 to 7 L min-1 led to a custom virtual impactor with a 1 µm cut point mounted on the PALMS instrument. The virtual impactor design closely followed Loo and Cork (1988) except it was slightly scaled for our flow and desired cut point. After the virtual impactor about 20 cm of 0.125 in. stainless steel followed by 20 cm of 0.125 in. conductive Teflon tubing led to the PALMS focusing inlet. The volumetric flow after the virtual impactor was 0.7 L min-1. The tubing was at aircraft cabin temperature except that about the last 5 cm was taped to a heat pipe connected to the PALMS ion source region, which was temperature controlled at 35 ∘C. This slight warming to >25 ∘C was done to avoid condensation in the aerodynamic focusing vacuum inlet on PALMS at low altitudes in the tropics. With the inlet tubing warmer than outside the aircraft, the relative humidity in the inlet was less than 40 % for the majority of boundary layer sampling so the water content of the sea-salt particles was reduced. Correlations between aerodynamic diameter and light scattering suggest that the salt particles did not effloresce in the inlet. There was probably insufficient residence time for efflorescence even when the relative humidity in the inlet was very low. For sea-salt aerosol mass concentrations we assume that the sea-salt particles were deliquesced, except if the outside relative humidity was less than 35 %, when we assume they were dry. At low temperatures sea salt partially effloresces at about 40 % relative humidity (Koop et al., 2000).

The University of Hawaii inlet on the DC-8 has been shown to quantitatively transmit particles as large as 3.1 µm, at least at low altitude (McNaughton et al., 2007). Both the focusing inlet on PALMS and the LAS inlet tubing do not transmit particles larger than about 4 or 5 µm, so results here represent particles smaller than about 3 µm (dry geometric diameter). Note that the measured size range of ∼0.18 to 3 µm often represents a minority of sea-salt aerosol by mass, as indicated by the size distribution of large sea-salt particles detected by the cloud probes in the cloud-free marine boundary. When comparing to other data or model results, the size range must be considered.

In-cloud data are excluded from the results shown here. Huebert et al. (1990) showed that up to 90 % of the largest sea-salt particles in the marine boundary layer can deposit to the walls of an inlet. Cloud droplets or ice crystals impacting a forward-facing aircraft inlet can act like a high-pressure washer to dislodge some of that salt, potentially leading to large sea-salt artifacts in clouds. During both ATom and previous missions (Murphy et al., 2004), PALMS observed anomalous particles in clouds, both sea salt and other compositions, reinforcing the decision to exclude in-cloud data.

Sea-salt particles were identified in the positive ion mass spectra using a combination of peaks, starting with a large Na+ peak (greater than at least 17 % or 30 % of the ion current from a particle, depending on other peaks in the spectrum). Identification also required a potassium peak of appropriate size. Particles were excluded as sea salt if they contained aluminum (from clay minerals), barium (from fly ash and other minerals), or a variety of other metals. A hierarchical cluster analysis (Murphy et al., 2003) was also used. Although the cluster analysis generated clusters of mass spectra that matched sea salt, criteria based on peak height were more reliable than the cluster analysis for identifying salt. Instead, the cluster analysis was most useful for excluding a few uncommon clusters that passed the peak height criteria but were not sea salt. For the great majority of sea-salt particles the identification was very clear. Figure 1 shows the mass spectrum of a single particle, chosen to be close to the average of all sea-salt mass spectra.

The main concern in identification is that at extremely low concentrations of sea-salt aerosol, as found over land or at high altitude, there may be a contribution from other particles, particularly dust from salt flats that is chemically similar to oceanic sea salt. A manual review was made of ∼100 particles automatically classified as sea salt and a similar number automatically classified as not sea salt. This was done for high-altitude coarse particles, the hardest region for the automatic classification. No definite sea-salt particles were missed and there was one definite false positive. More than 20 particles were borderline, mostly classified as sea salt by the automated algorithm. These were mostly spectra with sodium and potassium in a reasonable ratio for sea salt but magnesium at the wrong ratio. Note that if particles effloresce into inhomogeneous crystals, then the PALMS laser can sometimes preferentially ionize just one region of a particle with varying amounts of magnesium. There was a much smaller percentage of borderline identification in the marine boundary layer where higher humidity led to mostly liquid particles that ionize more uniformly. The Na+ peaks from sea-salt particles were often sufficiently large that they either saturated the detector or were broadened due to some sort of repulsion in the ion source. This tended not to affect the identification of a mass spectrum as sea salt but only the quantification of the Na peak.

Sea-salt particles can also be identified from negative ion spectra using chloride ions and cluster ions containing Na. Data in this paper are from positive ion spectra because in aged sea-salt particles the chloride can be almost entirely displaced by sulfate and nitrate, making identification more difficult with negative ions. In regions with fresh sea salt, results from negative ion spectra were nearly identical to those from positive ion spectra.

The number and mass of sea-salt particles were calculated for every 5 min of flight time in order to acquire enough mass spectra for a statistically significant normalization to the UHSAS and LAS size distributions. During climbs and descents, 5 min represents about 2.5 km in altitude. Mass spectra of sea-salt particles were typically acquired at a rate of more than one per second in the marine boundary layer and less than one per minute at high altitude.

Figure 2 shows concentrations of sea-salt aerosol at low altitudes measured by the PALMS and LAS combination compared to filter measurements of sodium (Dibb et al., 1999). The filter samples indicate more sea-salt mass, which is expected because the inlet to the filter sampler transmitted larger particles than the inlet to PALMS. The good correlation adds confidence to the PALMS measurements in the upper troposphere, which are much more sensitive than the filter sampler. Because PALMS samples each particle at a particular time, its composition can be associated with a particular altitude, aerosol concentration, water vapor concentration, and so forth. This allows for very long averaging times in similar air. For example, the average concentration of sea-salt aerosol in air with 10 to 20 ppmv of water can be calculated from collecting many such stretches of flight data even though they are not contiguous in time and might even be on different days. Furthermore, the PALMS single particle data are easier to screen for short periods of cloud and other artifacts than the extended filter samples are. One can eliminate short cloud encounters without losing the data before and after the cloud. The internal consistency of the data suggests that the detection limit for sea salt is better than 10 ng m-3 over a few minutes and better than 1 ng m-3 when hours of data are available. Exact detection limits depend on the size distribution and the amount of internal mixing. In favorable circumstances the detection limits can be ≪1 ng m-3.

The concentration of sea-salt aerosol in the marine boundary layer measured by PALMS was usually between 0.3 and 3 µg m-3. Such concentrations are reasonable considering the upper cut point of about 3 µm diameter. The concentrations in the marine boundary layer were highly variable and modestly positively correlated with the local wind speed at flight altitude. This is consistent with Shinozuka et al. (2004). Although the sea-salt aerosol production rate can be closely related to wind speed, the correlation with local wind speed is modest because the concentration also depends on the wind fetch and whether or not there has been recent precipitation. Sea-salt aerosol concentrations near the surface were also correlated with relative humidity. This is expected since low relative humidity would indicate that the air was not well mixed from the ocean surface up to the altitude of the DC-8. For example, during one boundary layer sampling leg over the ocean there was almost no sea salt at 160 m altitude, but the relative humidity was much less than 50 %, indicating limited surface influence for that particular leg.

There are already extensive measurements of sea-salt aerosol in the marine boundary layer (Lewis and Schwartz, 2004). The novel data here are the concentrations at higher altitudes. Figure 3 shows a latitude–altitude cross section of sea-salt aerosol concentrations over the Pacific Ocean. Even though they were in different seasons, systematic differences between ATom1 and ATom2 are not visible on a log color scale so they are combined in Fig. 3. More subtle differences depending on season are discussed below. The Pacific Ocean is shown because there was less mineral dust to complicate the analysis of very low concentrations; concentrations over the Atlantic Ocean were similar. Shinozuka et al. (2004) inferred sea-salt aerosol concentrations from nonvolatile particles. The results shown here agree that this was valid for their measurements at <2 km over the Pacific and Southern oceans. However, it is not true in general that most nonvolatile particles are sea salt. For example, at higher altitudes (e.g., 10 km) than Shinozuka et al. (2004) measured, nonvolatile particles would be many times more likely to be dust than sea salt, even over the middle of the ocean. Dust concentrations measured during ATom will be discussed in future publications.

A salient property of the distribution of sea-salt aerosol in Fig. 3 is that the concentration falls off rapidly with altitude, by about a factor of 10 for every 2 km. Above 6 km, the concentrations were almost always less than 10 ng m-3. The decrease at the top of the marine boundary layer was often very sharp in individual profiles. This was more apparent on past missions because the ATom mission ascents were relatively rapid. During the 2002 Intercontinental Transport and Chemical Transformation Experiment (ITCT) off the coast of California, PALMS was on the NOAA P-3 that often flew just above and below the top of the boundary layer. There, the sea-salt aerosol concentration could change by more than a factor of 10 in less than 100 m altitude at the top of the marine boundary layer. The few sea-salt particles in the upper troposphere were similar in size and more chemically processed than those in the marine boundary layer.

Figure 4 shows size distributions of sea-salt particles from PALMS during ATom1. Most of the sea-salt mass is in particles larger than 1 µm, most of the number is smaller than 1 µm. In the lower panel the size distribution from the upper troposphere is multiplied by 500 to get it on the same scale as the size distribution in the marine boundary layer, emphasizing the strong removal of sea salt. The number distribution is shown from the Southern Hemisphere (SH) because for submicron particles in the Northern Hemisphere (NH) both the number of particles and the fraction of sea salt are very steep functions of diameter, leading to large uncertainties when they are multiplied. The mass distributions and the Southern Hemisphere number distribution are less steep. The high and low altitude size distributions (Fig. 4b) show some differences of a factor of ∼3 as a function of diameter. Such detailed differences in the size distribution with altitude vary by region. We would emphasize instead that an overall large removal factor is present across the entire 0.3 to 3 µm diameter range. This may indicate that much of the removal of sea-salt aerosol was by nucleation scavenging in cloud rather than impaction by precipitation. Particles large enough to be measured by PALMS are all large enough to be condensation nuclei, whereas impaction scavenging is much more efficient for coarse particles than submicron particles (Pruppacher and Klett, 1997).

These are the first measurements of sea-salt aerosol over a wide range of altitudes and latitudes. Data are available from near the surface to about 12 km altitude from about 65∘ S to 80∘ N. The final ATom data set will add two more seasons and extend the data to beyond 80∘ S. Detailed comparisons with chemical transport models are underway (Bian et al., 2019; Zhang et al., 2019). One of these comparisons identified a problem in aerosol vertical transport in the CESM model (Yu et al., 2019).

Sea-salt aerosol's only source is at the surface and its only sink is by scavenging (i.e., sea-salt particles do not evaporate). That makes sea-salt aerosol a powerful tool to study wet removal of aerosol. The data here indicate that removal of sea-salt aerosol is very approximately proportional to the removal of water over a wide range of absolute humidity, with possibly more efficient removal in liquid water clouds than in ice clouds.