Measurements from the Department of Energy Atmospheric Radiation Measurement (DOE ARM) site on Ascension Island, Saint Helena (7.96696∘ S, 14.34981∘ W) during the Layered Atlantic Smoke Interactions with Clouds (LASIC) campaign are used to develop the sea spray mode retrieval. LASIC captured the annual and seasonal cycles of aerosol and cloud properties during an 18-month (April 2016–October 2017) deployment of the ARM Mobile Facility 1 (AMF1) (Miller et al., 2016; Zuidema et al., 2016). AMF1 measurements were collected at an isolated site on the windward flank of Green Mountain, away from the island's airport and other inhabited areas (Zhang and Zuidema, 2019). The prevailing wind direction measured by meteorological instrumentation during the campaign was 115 ± 10∘ (east-southeasterly), indicating persistent sampling of offshore maritime air. The ARM Mobile Aerosol Observing System, a component of AMF1, housed the instruments (described in the following subsections) and sampled from an inlet situated 10 m above ground level at an altitude of 365 m above sea level (Uin et al., 2019).

Episodic intrusions of airborne biomass-burning particles are carried into the Ascension Island marine boundary layer from South African wildfires annually during June–October (Zuidema et al., 2016). These events contrast sharply with the clean boundary layer that persists for the remainder of the year (November–May; Pennypacker et al., 2020). Nonmarine aerosol particles are limited during this “background” season, though a few transport events of African dust that entrain into the boundary layer have been documented as occurring during austral summer and fall months (January–April) in the southeast Atlantic (Kishcha et al., 2015; Virkkula et al., 2006). For background (non-biomass-burning) times without dust events, the aerosol population is expected to be largely from marine sources, of which sea spray represents a large fraction of the aerosol mass concentration. This work will focus on LASIC background season observations (November 2016–May 2017) (Table 1).

Nonmarine aerosol particles, specifically those from combustion sources, have been shown to influence the performance of sea spray mode retrievals by contributing number and mass concentrations that overlap with the region used to fit a mode in ambient size distributions (Modini et al., 2015; Saliba et al., 2019). Here we focused on periods when boundary-layer air masses are assumed to have a multiday marine history in order to reduce nonmarine sources and ensure that retrieval results are consistent with sea spray. In the case of LASIC observations, boundary-layer intrusions of biomass-burning aerosol can affect particle optical properties by increasing absorption and reducing scattering for sub- and supermicron particles (Delene and Ogren, 2002; Denjean et al., 2020). In particular, UHSAS instruments show undersizing of particles when highly absorbing biomass-burning aerosol is introduced into the particle population, due to heating from the instrument beam and subsequent particle vaporization and shrinkage (Howell et al., 2021). Although this effect is more likely to impact particles smaller than the assumed sea spray size (Dp<0.4 µm), any impact on potential submicron sea spray aerosol contributions would have an influence on the sea spray mode retrievals. To limit nonmarine influences on UHSAS-NEPH retrieval, we first isolated measurements during “clean marine” periods of the LASIC background season from November 2016 through May 2017 by applying the following criteria (Fig. 3):

CN3 concentration less than 600 cm-3, which was the approximate 90th percentile particle concentration during the LASIC campaign;

We additionally removed periods when the rain intensity exceeded 1 mm h-1 at any hour during the 2 h average to ensure a minimal influence of precipitation on the retrieval, namely wet scavenging of sea spray aerosol by rain droplets. 14 periods exceeded this rain intensity restriction. The combination of all criteria identified 909 2 h (nonraining) clean marine periods, which accounted for approximately 40 % of all available background season observations. The clean marine criteria reduced the average criteria values by 12 % for CN3, 7 % for CO, 70 % for rBC, and 15 % for SAE10 after applying these restrictions. The 909 available periods provided persistent marine conditions that included a low aerosol concentration (300 ± 90 cm-3) and a low combustion influence (15.7 ± 12.7 ng m-3) as well as a large scattering contribution from coarse particles (SAE10=0.66 ± 0.15).

A lookup table of scattering coefficients was developed by employing a modified Mie theory code based on the algorithm for red (700 nm), green (550 nm), and blue (450 nm) wavelengths (Bohren and Huffman, 1998; Mätzler, 2002). These wavelengths were chosen to match those used by the three-wavelength integrating nephelometer operated during LASIC (Sect. 2.1.2). Each of the three red, green, and blue (RGB) scattering coefficients is attributed to a combination of lognormal mode fitting parameters (Nt, Dg, and σg) that describe the shape of the sea spray mode. We use the canonical lognormal mode form to represent the number size distribution of the sea spray aerosol with the following equation (Seinfeld and Pandis, 2006): 1dNdlogDp=Nt2πlog⁡10σge-log⁡10Dp-log⁡10Dg22log⁡10σg2, where Nt is the number concentration of particles (cm-3), Dp is the particle diameter (µm), Dg is the geometric mean diameter (µm), and σg is the geometric standard deviation (or mode width; unitless). These values are 1–99 cm-3 for number concentration, 0.05–1.19 µm for mean diameter, and 1–4 for mode width, which provides over 157 000 possible sea spray mode solutions (Table 3). The range of fitting parameters was chosen to reflect those reported in laboratory experiments and field measurements (Table 6; Prather et al., 2013; Modini et al., 2010, 2015; Quinn et al., 2017; Saliba et al., 2019). Similar lookup tables have previously been used for optical measurement inversions (e.g., Lv et al., 2018; Veselovskii et al., 2002).

The scattering coefficients (bsca,MIE(λ)) are then related to the size distribution by integrating Eq. (2) over all particle diameters (0.01–10 µm), 2bsca,MIEλ=∫0∞πDp24Qsca(λ,m,Dp)dNdlogDpdlogDp. Qsca is the scattering efficiency, λ is the light wavelength, and m is the particle core refractive index (m=n+ik). To match the NEPH and UHSAS RH of 55 % during LASIC (Sect. 2.1.2), a constant m value of 1.45+i0 was selected to simulate sea spray particle scattering. This value is lower than the average refractive index reported for dry sea salt (real component = 1.5–1.6, imaginary component <10-6) (Wang and Rood, 2008; Randles et al., 2004; Bi et al., 2018) and was calculated as a mass-weighted mixture of salt with water, where water has a refractive index of 1.33 (Wang and Rood, 2008) (Sect. S2). We found no substantial variation in the retrieved fit parameters for the range of 1.4 + 0i to 1.6 + 0i, but correlations to sea spray tracers (chloride and wind speed) were lower for 1.40 + i0 (high RH) and 1.6 + i0 (low RH) (Sect. S2). The midrange value of m=1.45+i0 was used to approximate the scattering of sea salt for the measured RH of 55 %.

Since many solutions result from the constraints imposed by performing measurements at three wavelengths, we use the error between the nephelometer scattering and the Mie theory solutions to remove mode solutions that are not within the calculated error. The probability of solutions that meet the error threshold are then evaluated, and only the top 5 % most probable solutions are selected.

The measured supermicron scattering coefficient (bsca,1-10µmλ) is compared to scattering coefficients computed for each simulated sea spray size distribution and Mie theory value of bsca,MIE(λ) in the lookup table by calculating the absolute error at each wavelength (λ) using Eq. (3): 3Δbscaλ=bsca,1-10µmλ-bsca,MIEλ. The scattering error (Δbsca,RGB) is then computed by propagating the absolute error at each wavelength via Eq. (4): 4Δbsca,RGB=∑λ=RGBΔbscaλ2. The total sample space is then reduced by selecting solutions from the lookup table that fall below the error threshold (Δσsca,RGB) calculated for that measurement time, 5aΔbsca,RGB<Δσsca,RGB, where 5bΔσsca,RGB=∑λ=RGBσsca,1-10µm(λ)2+σsca,inst2. This error threshold incorporates the measured scattering variability at each wavelength for the averaging period (σsca,1-10µm(λ)) and accounts for the instrument error (σsca,inst, which is constant; Table 2). Figure 5 illustrates a time series of these error thresholds and the percent reduction of the Mie solution sample space (N=157 850) for each retrieval during the background season of LASIC. Using these error thresholds, the solution space is reduced by 98 % on average, with a reduction range of 83 %–99 %, resulting in approximately 1000–2000 possible solutions each time.

The majority of the solutions that are below the error threshold typically have a sea spray mode shape similar to those previously reported in the literature (Quinn et al., 2017; Saliba et al., 2019; Bates et al., 2012), namely mass mean diameters within or near the coarse-mode size range (1–10 µm) and mode widths of 2–3, decreasing to a number concentration of below 0.1 cm-3 before the 10 µm cutoff. However, some of the solutions that meet the error threshold criterion are either too wide or large relative to the reported range of sea spray modes (Table 6, Fig. S4), which is a limitation of having only three scattering wavelengths to constrain the mode. To remove the outlier solutions and to reduce the sample space to a more consistent group of solutions, we apply a restriction on the Mie solutions to consider only the most probable fitting parameters (NT, Dg, σg) based on their frequency of occurrence. To demonstrate this restriction, we consider the normalized probabilities of the fitting parameters from Mie solutions that fall below the scattering error threshold, Δσsca,RGB, for one retrieval during LASIC (Fig. 6a–c). This retrieval is selected as its Δσsca,RGB is within the average value for all LASIC observations during the clean marine background season (3.1 ± 2.1 Mm-1) and is representative of most cases assessed.

Mie solutions that meet the error threshold constrain the mode number concentration (NT) and mean diameter (Dg) of the sea spray mode, as shown by noting that the likely (probability >75 %) NT and Dg fall within narrow ranges of the low-error solutions (4 % and 10 % of the sample space, respectively), whereas there is a less constrained range for σg (45 % of the sample space) (Fig. 6a–c). Since σg has a wider range of probable values, this parameter is effectively constrained by considering the joint probabilities of NT|σg and Dg|σg. These probabilities are computed as 6PNT,σg=PNT|σg⋅Pσg 7PDg,σg=PDg|σg⋅Pσg. The highest joint probability values restrict solutions to localized regions of the solution space (Fig. 6d–e). Selecting only the upper 5th normalized probability percentile of joint probabilities further restricts the full Mie solution sample space by an additional 16 % and 13 % for NT|σg and Dg|σg, respectively (Fig. 6d, e).

To assess which of these joint probability restrictions provided the most realistic sea spray mode results, retrieved mass concentrations and fitting parameters during the clean marine background season of LASIC were compared (Fig. S5). Restricting low-error solutions using the Dg|σg joint probability led to retrieved sea spray mass concentrations that were 20 %–30 % higher than the NT|σg combination, had mass mean diameters smaller than NT|σg (0.68 ± 0.08 µm vs. 1.47 ± 0.17 µm), and had much broader mode widths (3.8 ± 0.2 vs. 2.4 ± 0.3). Although the mass mean diameters retrieved using the Dg|σg joint probability were within the range of reported values from the literature (Table 6; Quinn et al., 2017; Saliba et al., 2019; Bates et al., 2012), the exceptionally broad widths outside of the reported range suggest that restricting using Dg|σg does not effectively constrain the mode width as well as the NT|σg combination. For these reasons, only sea spray modes with fitting parameters in the top 5th percentile of the NT|σg joint probability are used for this method, which typically results in reduction of the full sample space to 300–500 Mie solutions for each measurement time.

Uncertainty and variability have been shown to impact the assumed size distributions and optical properties of aerosols when using inversion techniques (Viskari et al., 2012; Frie and Bahreini, 2021). Here, instrument uncertainty and measurement variability are incorporated into the fitting method by introducing random noise into the size distribution based on the size (σD) and number concentration (σPNSD,meas (Dp), σPNSD,inst) uncertainties (Sect. 2.1.3; Table 2). Perturbations are simultaneously made to the measured number size distribution and particle diameters. Each bin of the size distribution is perturbed by introducing Gaussian noise that samples a random number from a normal distribution. The sampled normal distribution is defined by a mean (μ) and a standard deviation (σ), where μ is the measured PNSD averaged over the time interval (σPNSD,meas (Dp)) and σ is the combination of the errors from temporal variability and instrument concentration uncertainty (which is constant): 8σ=σPNSDDp2+σPNSDinst.2. The diameters in the size distributions are perturbed by shifting the size bins by the same value with a μ of 0 and σ as the instrument size uncertainty (σD). Size distributions are perturbed 100 times to provide a sample space of Nperturb+1 (the measured size distribution and 100 perturbations). Each probable Mie solution retrieved in Sect. 3.2 is then tested for the Nperturb+1 cases within the fitting region of the measured submicron size distribution described in Sect. 3.4.

The last step in retrieving sea spray modal properties is to select Mie solutions that most closely match the shape of the measured size distribution in a specified fitting region. Given the correlation of supermicron scattering and sea spray mass concentration during clean marine conditions (Chamaillard et al., 2006; Kleefeld et al., 2002; Quinn et al., 1998), all PNSDs are converted to particle mass size distributions (PMSDs) using 9dMdlogDp=π6ρDp3dNdlogDp, with the assumptions of spherical particle homogeneity and constant sea spray density ρ. UHSAS and NEPH measurements were collected at 55 % RH, so 1.3 g cm-3 is used as the estimated sea spray particle density by calculating a mass-weighted mixture of salt with water (Sect. S3).

To account for the consistent UHSAS artifacts at 0.6 and 0.85 µm (Sect. 2.1.1), we restrict the measured size distributions used to fit the Mie-theory-simulated size distributions for diameters larger than 0.4 µm (0.4 to 1 µm UHSAS range) to 0.38–0.83 µm (the closest UHSAS diameter size bins within the specified range), which weights the comparison toward smaller sizes and effectively reduces the influence of the largest artifact while maintaining the shape of the accumulation mode “shoulder” (Fig. 1). For each Mie solution (PMSDMIE), a residual sum of squares is computed between the measured and perturbed PMSDs as 10FitRSSj,k=∑PMSDD,j-PMSDMIED,k2, where the index j represents each measured or perturbed size distribution and k is the index of the low-error Mie solution. The fit RSS provides a quantifiable metric for comparing the scattering-retrieved mode to the expected sea spray fitting region. Chi-square “goodness of fit” calculations were also performed and provided similar retrieved sea spray modal properties to those retrieved using the fit RSS minimization. The high-probability Mie solution with the minimum fit RSS for the measured and noise-perturbed size distributions is chosen to establish a range of mode fits. This Monte Carlo approach allows for the selection of Mie solutions that capture the shape of the accumulation mode “shoulder” given the uncertainties associated with measurement variability and instrument error, and it provides a statistically robust sample space for retrieving a unique sea spray mode solution. From the range of fit-RSS-minimized sea spray modes, 95 % confidence intervals of each fitting parameter are calculated to further constrain the most probable solution that fits the measured size distribution in the expected sea spray size range. On average, 30–40 solutions remain from this perturbation analysis for each measurement time, with variabilities of 4 % in number concentration, 3 % in geometric mean diameter, and 1 % in geometric standard deviation, based on the sample means and upper and lower limits of the 95 % confidence intervals. The low variabilities of these fitting parameters demonstrate consistent mode retrievals within the perturbations and stability of the retrieval procedure. Lastly, the fitting parameter solutions that are both highly probable (Sect. 3.2) and within the 95 % confidence interval of fit-RSS-minimized modes are averaged to produce a single lognormal sea spray mode with the uncertainty of the solution defined as the upper and lower bounds of the 95 % confidence interval.

We applied the SEMS-APS fitting algorithm described in Sect. 4 to measured UHSAS number size distributions (hereafter identified as “UHSAS-only”) and compared sea spray mode results with those retrieved using UHSAS-NEPH. Sanchez et al. (2021) have recently applied the algorithm to submicron UHSAS size distributions obtained from aircraft measurements in the marine boundary layer of the Southern Ocean and found it to be a good approximation of the sea spray contribution to the CCN number concentration by comparison to quantified sub- and supermicron sea salt particles using electron microscopy. The mode fitting parameters NT and Dg of the UHSAS-only method were converted to mass-derived values for comparison with UHSAS-NEPH using Eq. (7) for the same particle diameter range (Dp=0.01–10 µm). UHSAS-only sea spray modes were fitted for 90 % of the 2 h average size distributions during the clean marine background season. When fits could not be achieved, there was generally noise in the measured size distribution or other common singularities (see the supplement of Saliba et al., 2019). Summary statistics comparing parameters retrieved using UHSAS-NEPH and UHSAS-only are provided in Table 5.

Integrated sea spray mode mass concentrations ranged from 0.008 to 23.5 µg m-3 employing UHSAS-only and from 0.18 to 23.0 µg m-3 with UHSAS-NEPH. Average sea spray mass concentrations of 1.3±2.2 µg m-3 and 8.37 ± 4.1 µg m-3 were observed for UHSAS-only and UHSAS-NEPH, respectively, which indicates that using UHSAS-only provides lower sea spray mass retrievals because UHSAS-only fits are constrained solely by the accumulation mode shoulder. Sea spray mass concentrations from both methods exhibit generally consistent concentrations during the clean marine periods of LASIC with no apparent seasonality (Fig. 9a), but there are distinct differences in the retrieved mode diameters and widths (Fig. 9b, c).

Sea spray mode retrievals using UHSAS-only were much smaller in mass mean diameter than those of UHSAS-NEPH, with averages of 0.68 ± 0.01 µm and 1.47 ± 0.17 µm, respectively (Fig. 9b). The range of mass mean diameters was similar for both retrieval methods, 0.4 to 1.6 µm for UHSAS-only and 0.6 to 1.9 µm for UHSAS-NEPH, though the majority of the UHSAS-only mean diameters were submicron (Fig. 9b). Just over 2 % (14) of the UHSAS-only retrievals had mass mean diameters in the coarse mode (>1 µm), compared to 91 % (723) of those of UHSAS-NEPH. The peak in mass mean diameters at sizes within the coarse mode using UHSAS-NEPH can be attributed to the additional contributions of supermicron mass identified by the nephelometer supermicron scattering that are not constrained by the UHSAS submicron size distribution. These predominately supermicron mean diameters are consistent with the assumption that sea salt particles contribute a large amount of mass and scattering within the coarse mode, which is simply not captured by the submicron UHSAS distributions alone.

On average, UHSAS-only retrievals were narrower at 1.8 ± 0.4 than those retrieved with UHSAS-NEPH at 2.4 ± 0.3 (Fig. 9c). The range of mode widths using UHSAS-only varied from particularly narrow (1.3) to a fairly broad and unconstrained width of 5.3, compared to 1.1 to 3.97 for UHSAS-NEPH. UHSAS-only widths are determined solely by the shape of the large accumulation mode shoulder in the UHSAS number size distribution and include a variety of widths based on how flat or sharp the slope of this shoulder may be. Mode widths retrieved from both methods were predominately narrower than a value of 3, with only 2 % of UHSAS-only modes and 1 % of UHSAS-NEPH modes greater than this value. The average UHSAS-only mode narrowness again reflects the absence of supermicron size distribution measurements. Sanchez et al. (2021) reported a similarly narrow average mode width (1.44 ± 0.25) for marine boundary layer sea spray aerosol retrieved with UHSAS-only. Conversely, Yu et al. (2019) reported a broad mode width (geometric standard deviation of 2.7) and a volume peak at a supermicron diameter (approximately 2 µm) for 24 h averaged sea salt particle size distributions at Ascension Island using Aerosol Robotic Network retrievals, which is more consistent with UHSAS-NEPH observations. These results indicate that UHSAS-only may provide a good estimate of the sea spray number concentration, which predominantly consists of submicron-sized particles (Sanchez et al., 2021), but the lack of supermicron measurements makes it unable to adequately identify mass contributions from larger particles.

Comparing the sea spray mode fitting parameters to those found in the literature shows that retrieved modal properties for both methods are within the ranges of reported values of mass mean dry diameter (0.25–1.6 µm) and mode width (1.4–3) (Fig. 9b, c and Table 6), while it should be noted that these LASIC retrievals are for 55% RH. The median reported mass diameter (0.88 µm) falls in between the statistical mode values (peaks in histograms) of the retrieval methods at the upper end of UHSAS-only and lower end of UHSAS-NEPH (Fig. 9b), showing consistency with other ambient sea spray mode measurements (Quinn et al., 2017; Sanchez et al., 2021; Modini et al., 2015). Retrievals using only the UHSAS number size distributions generally show better agreement in terms of mode size with laboratory-based bubble bursting and breaking wave flume studies, which had mass mean diameters that were less than 1 µm (Prather et al., 2013; Bates et al., 2012; Modini et al., 2010). UHSAS-NEPH retrievals are more consistent with field observations of the sea spray size distribution across several ocean basins, including open ocean studies in the Pacific and Atlantic, in which measurements of supermicron size distributions were incorporated (Quinn et al., 2017; Saliba et al., 2019; Modini et al., 2015). The UHSAS-NEPH retrieval of sea spray mode width was at the lower end of the reported laboratory and field measurement values, while the narrow UHSAS-only modes were generally outside the spread of the majority of reported values. These differences in the ranges of the retrieved values show that the mode width is the least constrained parameter derived by UHSAS-only and UHSAS-NEPH, although the scattering-based constraint provides some apparent improvement compared to UHSAS-only when supermicron mass contributions are considered.

Figure 10 displays six UHSAS-NEPH sea spray mode fits that are characteristic of the retrieval procedure. Cases with different modal properties (diameter, width, mass) and a comparison of this retrieval with the UHSAS-only algorithm are presented. For both methods, the accumulation mode (0.4–1 µm) is generally well characterized by the retrieved modes, which follow the shape of this broad shoulder closely. Differences between the UHSAS-only and UHSAS-NEPH retrieved mass size distributions become more apparent just before the 1 µm size limit of the UHSAS. For narrow modes (σg∼2), the UHSAS-only method appears to generally be sufficient to quantify the sea spray mode mass concentration in the absence of supermicron scattering measurements (Fig. 10a). In these cases, the lower contribution of coarse particles measured by the nephelometer supermicron scattering adds little information at the tail of the size distribution. The limitations of fitting a sea spray mode based solely on the shape of the accumulation mode “shoulder” in the number size distribution are illustrated in Fig. 10b, c, d. The broadness of the shoulder at submicron sizes (0.38–0.83 µm) forces the UHSAS-only retrieval to include more particles from submicron sizes and fewer from the supermicron regime. This leads to lower mass concentrations in UHSAS-only compared to UHSAS-NEPH. Mode retrievals using only the UHSAS size distribution likely underestimate much of the mass at supermicron sizes, as seen by the sharper tailing off of the UHSAS-only modes in the coarse regime (Fig. 10b, c, e, f), with up to 5 µg m-3 of sea spray mass difference in the cases considered here.

We next compared sea spray mode mass concentrations from UHSAS-only and UHSAS-NEPH with available sea spray tracers to evaluate the extent to which the retrieved modal properties represent realistic sea spray size distributions.

The submicron mass concentration measured by the ACSM provided a trace chloride signal that could be used to examine sea salt mass concentrations from the UHSAS-only and UHSAS-NEPH retrievals. Comparing the retrieved submicron sea spray mass with the ACSM chloride mass, we found correlations of R=0.39 (p<0.05) for UHSAS-only and R=0.35 (p<0.05) for UHSAS-NEPH (Fig. 11a, b). While the ACSM signal is a very indirect chemical measurement of refractory chloride, these positive correlations provide support for the ability of both methods to identify retrieved modes as sea spray. The higher correlation of UHSAS-only over UHSAS-NEPH may be a result of the submicron sampling range of the ACSM, the diameter range in which UHSAS-only is solely constrained. Additionally, much of the sea spray mass retrieved from UHSAS-NEPH is concentrated in the supermicron regime, which is not observed in ACSM submicron chloride measurements.

Correlations between sea spray mass and the supermicron scattering coefficient were assessed for UHSAS-only and UHSAS-NEPH using retrievals that followed the criteria discussed in previous sections. Sea spray mass concentrations from UHSAS-only were consistently less than 2 µ g m-3 and were not able to quantitatively explain the variability in the measured supermicron scattering at 550 nm, as indicated by a very weak correlation (R=0.01, p<0.05; Fig. 11c). Although the stronger correlation between sea spray mass concentration and scattering obtained with UHSAS-NEPH (R=0.84, p<0.05, Fig. 11d) is an expected result given that the retrieved modes are constrained by scattering measurements, the lack of correlation from UHSAS-only emphasizes the importance of including supermicron particle size measurements to adequately characterize how optical properties are influenced by the sea spray size distribution.

The functional relationship between sea spray mass concentration and wind speed was evaluated using linear regression on the UHSAS-only and UHSAS-NEPH retrievals, which resulted in correlation coefficients of 0.1 and 0.2, respectively (p<0.05) (Fig. 11e, f). These correlations were lower than the values (in the range of 0.4 to 0.9) reported for numerous basins of the global ocean (Liu et al., 2021; Russell et al., 2010; Feng et al., 2017; Saliba et al., 2019). The calm and generally invariant wind speed observed during the LASIC background season (7.1 ± 1.4 m s-1) could explain the poor correlations for both methods, as the lack of dynamic range in wind speed at Ascension Island reduces the degree to which it explains the variability in concentration. The variability may instead be impacted by marine airmass transport of sea spray aerosol that is from source regions away from Ascension and reaches the island, rather than local wind conditions, which would not be resolved in the correlation (Grythe et al., 2014). The improved correlation of UHSAS-NEPH in comparison to UHSAS-only shows some added value of applying the supermicron scattering constraint to characterize the sea spray mode. Evaluation of the method using North Atlantic measurements (Sect. 4), where more dynamic wind conditions were observed, shows that the use of the supermicron scattering constraint could be comparable to the constraint from measured supermicron size distributions when estimating sea spray production from wind speed, as has been done in previous work (Saliba et al., 2019; Modini et al., 2015; Quinn et al., 2017). These results indicate that the incorporation of nephelometer scattering as a constraint for supermicron particle size provides a reasonable replacement for measured supermicron mass size distributions.