Portal de Periódicos da CAPES

Modeled optical thickness of sea-salt aerosol

2011; American Geophysical Union; Volume: 116; Issue: D8 Linguagem: Inglês

10.1029/2010jd014691

2156-2202

W. L. Madry, O. B. Toon, Colin O’Dowd,

Meteorological Phenomena and Simulations

Journal of Geophysical Research: AtmospheresVolume 116, Issue D8 Aerosol and CloudsFree Access Modeled optical thickness of sea-salt aerosol William L. Madry, William L. Madry [email protected] Department of Atmospheric and Oceanic Sciences and Laboratory for Atmospheric and Space Physics, University of Colorado at Boulder, Boulder, Colorado, USA Now at Department of Chemical Engineering, Bucknell University, Lewisburg, Pennsylvania, USA.Search for more papers by this authorOwen B. Toon, Owen B. Toon Department of Atmospheric and Oceanic Sciences and Laboratory for Atmospheric and Space Physics, University of Colorado at Boulder, Boulder, Colorado, USASearch for more papers by this authorC. D. O'Dowd, C. D. O'Dowd School of Physics and Center for Climate and Air Pollution Studies, Environmental Change Institute, National University of Ireland, Galway, Galway, IrelandSearch for more papers by this author William L. Madry, William L. Madry [email protected] Department of Atmospheric and Oceanic Sciences and Laboratory for Atmospheric and Space Physics, University of Colorado at Boulder, Boulder, Colorado, USA Now at Department of Chemical Engineering, Bucknell University, Lewisburg, Pennsylvania, USA.Search for more papers by this authorOwen B. Toon, Owen B. Toon Department of Atmospheric and Oceanic Sciences and Laboratory for Atmospheric and Space Physics, University of Colorado at Boulder, Boulder, Colorado, USASearch for more papers by this authorC. D. O'Dowd, C. D. O'Dowd School of Physics and Center for Climate and Air Pollution Studies, Environmental Change Institute, National University of Ireland, Galway, Galway, IrelandSearch for more papers by this author First published: 26 April 2011 https://doi.org/10.1029/2010JD014691Citations: 14AboutSectionsPDF ToolsRequest permissionExport citationAdd to favoritesTrack citation ShareShare Give accessShare full text accessShare full-text accessPlease review our Terms and Conditions of Use and check box below to share full-text version of article.I have read and accept the Wiley Online Library Terms and Conditions of UseShareable LinkUse the link below to share a full-text version of this article with your friends and colleagues. Learn more.Copy URL Share a linkShare onFacebookTwitterLinkedInRedditWechat Abstract [1] We simulate the generation and microphysical evolution of sea-salt aerosol using a climatologically driven 3-D microphysical model for the year 2006. We then apply Mie theory to calculate the extinction and scattering efficiencies of our transported, size-resolved sea-salt aerosol, accounting for hygroscopic growth due to changes in ambient relative humidity. We calculate the column optical thickness of our modeled sea-salt aerosol for comparison to three previously published wind speed–dependent clean marine air optical thickness formulations. Variously derived from optical thickness measurements and retrievals taken from the Midway Island AERONET site, the satellite-based MODIS instruments, and the Global Atmospheric Watch (GAW) site at Mace Head, Ireland, the three formulations report similar background levels of clean marine AOT at zero wind speed but significantly different functional dependencies for nonzero wind speeds. We find that our modeled sea-salt aerosol optical thickness very closely depends on the square of surface wind speed under steady state conditions. This relationship is consistent across all latitudes. However, due to the fact that steady state winds are seldom maintained, the 24 h mean wind is more frequently applicable to calculations of sea-salt AOT, with only slightly diminished accuracy. Key Points Optical thickness of SSA goes with square of surface wind speed This relationship is latitude independent Both steady state and average wind speeds are useful for this calculation 1. Introduction [2] Ubiquitous in nature, the sea-salt aerosol has been modeled extensively in recent years [Fairall et al., 1994; Gong, 2003; Vignati et al., 2004; Clarke et al., 2006; Mahowald et al., 2006; Zakey et al., 2008]. Despite the efforts of many in the field, successfully modeling sea-salt generation, transport, and removal has remained an elusive goal. Lewis and Schwartz [2004, Figure 44] show a scatterplot of modeled versus measured sea-salt mass for several different models, which generally have poor correlation between model and data. [3] While salt mass is important as a constraint for sea-salt models, the radiative properties of sea-salt aerosol are equally important to understand. Since the mass and optical thickness are controlled by different parts of the size distribution, simulating one does not guarantee a good simulation of the other. The column optical thickness of sea salt is generally lower than that of other commonly studied aerosols such as black carbon, dust, and volcanic ash, as the column burden of such aerosols is high near their sources. However, given the oceanic expanse of the sea-salt aerosol source region and the findings of Loeb and Kato [2002], among others, that the direct radiative effect of sea-salt aerosol over the ocean is negative and on the order of a few W/m2, sea salt's influence on the planetary energy budget may be quite extensive. In addition to scattering visible and near infrared sunlight, sea-salt droplets act as cloud condensation nuclei [Campuzano-Jost et al., 2003]. Clouds reflect incoming solar radiation and trap outgoing planetary radiation. As such, any influence on clouds by sea-salt aerosol represents an indirect radiative effect [Twomey, 1974, 1991; O'Dowd et al., 1999], which may interact with the indirect effects triggered by anthropogenic aerosols. Thus, an improved understanding of the role of sea-salt aerosol in the atmosphere is needed to improve our understanding of global climate. [4] In this paper, we simulate optical thickness (AOT) of the sea-salt aerosol for the year 2006. Numerous groups have previously simulated sea-salt aerosol optical thickness. However, it is very difficult to find data that can be used to verify these simulations. The total AOT for clean marine air masses seldom exceeds 0.1. Other aerosols, including dust and non-sea-salt sulfate, can easily dominate AOT at even the most remote locations. Several research groups have attempted to derive a relationship between surface wind speed and clean marine AOT using data from various sources. They find that sea-salt aerosol optical thickness can be identified from properly selected and filtered optical thickness measurements. This technique works because sea-salt aerosol is often produced locally by the action of wind stress on the ocean surface. Because wind power goes with the cube of wind speed, sea-salt aerosol flux into the atmosphere has been parameterized in models as a nearly cubic function of wind speed [Monahan, 1986; Gong, 2003] As such, sea-salt flux is very sensitive to changes in surface wind speeds. These derived relationships may provide a useful test for models of sea-salt aerosol, because the observed relations between wind speed and sea-salt optical thickness seem to be globally representative. Here we test to see if our model does reproduce the observed wind speed–optical depth relations, and whether the same relationship applies across the world's oceans. [5] In this paper, we report on our efforts to compare our modeled sea-salt aerosol optical thickness to the wind speed–dependent formulas for clean marine aerosol optical thickness data based on relationships provided by Smirnov et al. [2003], Satheesh et al. [2006], Mulcahy et al. [2008], Glantz et al. [2009], and Huang et al. [2009]. 2. Aerosol Model Description [6] The CARMA model simulates the advection, diffusion, sedimentation, coagulation, and condensational growth of atmospheric aerosols [Toon et al., 1988]. CARMA has been used to study a variety of aerosol problems, including dust storms and smoke plumes [Westphal et al., 1988; Colarco et al., 2002, 2003a, 2003b; Matichuk et al., 2007, 2008]. Dust storms are particularly relevant, because like sea salt, lifting rates are a strong function of wind speed. These dust simulations have been compared to data from actual events, leading to useful feedback on the success of the model. To simulate actual events, we use CARMA in conjunction with the Model for Atmospheric Transport and Chemistry (MATCH), a National Center for Atmospheric Research (NCAR) model [Rasch et al., 1997], to recreate the state of the atmosphere for a given time period from archived meteorological data. This same meteorological data drives aerosol source and deposition fluxes, advection and diffusion of aerosols, and other microphysical processes including rain-out. The model output is then analyzed for comparison to measurements of marine AOT and relationships between wind speed and marine AOT that have been described previously. 3. Dynamics, Transport, and Microphysics [7] The transport module of CARMA is driven by meteorological fields produced by MATCH [Rasch et al., 1997], which in turn accepts data from the National Center for Environmental Prediction (NCEP) reanalysis project [Rasch et al., 1997]. The NCEP reanalyses for the model period are fed into MATCH at 6 h intervals, available each day at 0000, 0600, 1200, and 1800 UTC. MATCH interpolates atmospheric conditions including temperature, humidity, and wind vector at thirty minute intervals on a 1.875° × 1.875° grid, matching the spatial resolution of the NCEP reanalyses. In addition to interpolating reported values from the NCEP reanalyses, MATCH calculates cloud fields and produces diffusion coefficients for the planetary boundary layer. The global coverage from MATCH extends to 35 km in height, divided into 28 vertical sigma layers. One issue we find to be important is the limited boundary layer resolution of MATCH. For instance, the first model layer is about 100 m thick. Large sea-salt particles can have a significant gradient over such thicknesses, as discussed by de Leeuw [1986] and Hoppel et al. [2002, 2005]. We archive the output from MATCH for use as a driver for CARMA, which runs at the same spatial and temporal resolution. [8] For the simulations described in this paper, the model domain wraps completely around the earth and covers a latitude range of 80°S to 80°N, extending vertically through all 28 sigma layers. Because we are working in sigma coordinates, the surface pressure of the model domain varies with meteorological and topographic conditions. Additionally, while the sigma thickness of individual layers is constant over the model domain, the geometric thickness of individual grid cells varies spatially and temporally. The sigma thickness varies from layer to layer. There are about six layers in the lowest kilometer of the model. [9] Advection and diffusion are calculated in CARMA using a flux form scheme for the outer advection operator and an advective form scheme for the inner operator following Lin and Rood [1996]. The flux form scheme employs the Piecewise Parabolic Method (PPM) [Colella and Woodward, 1984] and the advective form scheme employs a first-order semi-Lagrangian technique [Bates and McDonald, 1982]. The PPM allows a fully implicit solution for Courant number greater than 1 (Cr = v * dt/dz, where v is the advection velocity in the direction dz, dt is the length of our model time step, and dz is the linear size of a grid box in the appropriate dimension). Additionally, turbulent mixing in the planetary boundary layer employs a scheme that provides a countergradient transport term. Under unstable and convective conditions, the size of turbulent eddies approaches the depth of the planetary boundary layer [Hack, 1994; Zhang and McFarlane, 1995]. [10] Sea-salt aerosol is hydrophilic. For the simulation described in this paper, we keep track of the dry sea-salt aerosol number in each of 16 (dry) radius bins. The 16 radius bins are volume centered and span the range 1.45 × 10−2μm to 50.5 μm. This results in smallest bin lower boundary of 0.01 μm and a largest bin upper boundary of 60 μm. Following Fitzgerald [1975], we calculate an appropriate wet radius, rwet, for each bin every time step, from which we calculate the size-dependent fall and deposition velocities of our transported sea-salt aerosol as discussed below. This method assumes that the sea-salt particles remain in equilibrium with ambient relative humidity. A potential error in this assumption is that large (rdry > 10 μm) salt particles may actually require a significant period of time to reach equilibrium with water vapor. However, such particles have short atmospheric lifetimes (hours) relative to our regional transport times (days). Another potential error in our growth calculations arises if the sea-salt particles effloresce, which occurs if the ambient RH drops below ∼45% [Tang et al., 1997]. Because of hysteretic effects, sea-salt particle size depends not only on the current ambient RH, but also on the history of RH to which the particle has been exposed [Junge, 1952; Richardson and Spann, 1984; Cohen et al., 1987]. However, as the marine boundary layer is seldom, if ever, dry enough for sea-salt particles to effloresce, we make no accommodation for this possibility. [11] Particle fall velocities are calculated by treating all particles as spheres, which is a reasonable assumption for wet particles [Pruppacher and Beard, 1970]. Fall velocities are added to the vertical wind velocities for each radius bin. In the lowest layer, we calculate a wet particle deposition velocity from sedimentation, diffusion, and turbulent deposition to the surface [Seinfeld and Pandis, 1998], following the two-layer method described by Shao [2000]. [12] To calculate aerosol wet removal rates, CARMA examines the cloud fields and precipitation rates produced by MATCH. In a somewhat simplistic fashion, CARMA employs a size-independent collection efficiency for the below-cloud scavenging following Dana and Hales [1976] and Guelle et al. [2001] to calculate the amount of aerosol removed from a model grid box in a given time step. We assume two number-lognormal raindrop size populations, for stratiform and convective precipitation, with rmed (the mode radius) of 0.02 and 0.1 cm, respectively, and σ = 2.0 for both [Yau and Rogers, 1984]. Under this scheme, the below-cloud convective scavenging is less efficient per unit of precipitation than the stratiform scavenging. In-cloud scavenging is handled by assuming a size-independent collection efficiency of 10% per time step in grid boxes containing a stratiform cloud layer. In convective clouds, we determine scavenging rates by calculating the fraction of a grid box that is swept by precipitation, using the cloud volume and precipitation rates given by MATCH, and remove all aerosol in that fraction. 4. Aerosol Optical Thickness Calculations [13] We calculate the column optical thickness of our modeled sea-salt aerosol by treating the sea-salt particles as spheres and applying a Mie code [Wiscombe, 1979, 1980] to calculate Qext, the extinction efficiency of the sea-salt aerosol particles, as a function of dry particle bin size and ambient relative humidity. We assume that the refractive index of the wet droplets varies with water uptake, which increases with increasing relative humidity. The refractive index for the wet sea-salt particles is calculated as the volume-weighted average of the refractive indices of dry sea salt (n = 1.50–1.55e-8i) and pure water (n = 1.33), using the Fitzgerald [1975] swelling scheme as described above to calculate the wet particle radius, rwet. Column optical thickness is then calculated as the integrated sum of particle concentration multiplied by Cext, the single particle extinction coefficient in each bin. The single particle extinction coefficient is given by the relation Cext = πrwet2Qext. In Figure 1a, we show the 870 nm extinction efficiency of our modeled sea-salt aerosol as a function of dry particle size and relative humidity. The smallest particle sizes have a negligible extinction efficiency, while the largest particles have an extinction coefficient that approaches the large particle limit of 2. Particles in the size range of 0.2 to 1 μm have the highest extinction efficiencies in our model. In Figure 1b, we show a typical area size distribution of our transported sea-salt aerosol. Figure 1c shows the calculated size-dependent particle residence times, where lifetime τ is calculated for each size bin by dividing total atmospheric sea-salt aerosol mass, M, by mass production rate, P, in the model at every time step. Large particles, which dominate the injected mass, are removed quickly and contribute little to the surface area of the particles. The peak in the area distribution occurs at 1 μm dry radius, and particles in the size range 0.3–2.0 μm dry radius contribute the bulk of the calculated aerosol optical thickness in our model (Figure 1d) because of their high extinction efficiency and concentration. Figure 1aOpen in figure viewerPowerPoint Sea-salt aerosol particle extinction efficiency as a function of dry aerosol radius for four relative humidity conditions. The peak in extinction efficiency is seen for submicron-sized particles, while larger particles tend toward the limit of Qext = 2. Figure 1bOpen in figure viewerPowerPoint A typical area size distribution of our modeled sea-salt aerosol, which shows a peak in the range of 1 μm. Figure 1cOpen in figure viewerPowerPoint Aerosol lifetime for the particles modeled in this study. Clearly, given the size-dependent extinction efficiencies shown in Figure 1a, the area size distribution shown in Figure 1b, and the size-dependent particle residence times in Figure 1c, sea-salt aerosol optical thickness is dominated by particles in the size range 0.03–0.9 μm. Figure 1dOpen in figure viewerPowerPoint Modeled sea-salt aerosol optical thickness at 870 nm, per model size bin. The values are averaged over the entire modeled ocean domain, time averaged for a 2 week period. 5. Sea-Salt Source Function [14] A wide variety of sea-salt source functions developed for use in aerosol models have been extensively examined and reported in recent years [Andreas, 2002; Lewis and Schwartz, 2004; Clarke et al., 2006; O'Dowd et al., 2004]. For this experiment, we use the sea-salt source function originally developed by Monahan [1986] and revised by Gong [2003], with further refinement as discussed below. We chose to use the Gong [2003] source function because it is in common use, however we modified it to use the surface stress as an input wind rather than the 10 m wind as discussed below. [15] The sea-salt source function given by Monahan [1986] relies on the combination of two separately quantified processes, oceanic whitecap coverage and bubble bursting, to estimate the production of sea-salt aerosol at the ocean-atmosphere interface. The first process, whitecap coverage from breaking waves, depends on surface level winds. Wind stress acts on the ocean surface to create swells. Wind speeds above a given threshold, approximately 4 m/s, generate waves that break over and entrain air into the surface layer of the ocean. In the wake of such waves, the entrained air returns to the ocean surface as tiny bubbles, creating whitecap foam. In a secondary process, the individual bubbles in the foam subsequently burst, emitting sea-salt aerosol droplets from the film and central jet of each collapsing bubble. The size distribution of the emitted sea-salt aerosol depends on the size distribution of the source bubbles that burst at the ocean-atmosphere interface. This process has been studied using high-speed photography in wave tank experiments [Resch et al., 1986; Resch and Afeti, 1991, 1992; Spiel, 1998]. Monahan [1986] combines these two processes, wave breaking and bubble bursting, into a wind speed and radius-dependent sea-salt source function suitable for use in models with surface wind speeds (U10) up to 15 m/s and sea-salt particles sized 0.8–8 μm radius under conditions of 80% RH (r80), and Andreas [2002] and Gong [2003] extend this size range to 0.01–12 μm. [16] More recent work [O'Dowd and Smith, 1993; Clarke et al., 2003] demonstrates that a function describing the source flux of sea-salt aerosol must consider particles much smaller than originally described by Monahan [1986]. Data from radial differential mobility analysis (RDMA) instruments detect sea-salt particles to a lower size limit of 0.01 μm dry diameter (Ddry) [Clarke et al., 2003], and even smaller sea-salt particles may exist. Gong [2003] extends the range of the sea-salt source function down to r80 = 0.01 μm, roughly equivalent to the dry diameter of 0.01 μm that has been observed. We further enhance the sea-salt source function given by Gong [2003] by using surface stress, rather than 10 m wind speed, as the driving physical parameter. Additionally, we account for subgrid-scale variability in surface stress by applying a Weibull probability distribution to the surface stress used in the MATCH/CARMA model. Finally, we apply the surface source correction term given by Hoppel et al. [2005] to modulate the flux of sea-salt particles into the lowest layer of our model. As such, the sea-salt source function used in this numerical experiment is given by: where dF/dr80 is the number flux of sea-salt particles of radius r80 [μm] in #/m2/μm/s, given height and roughness lengths of z and z0, von Karman constant κ, surface stress τ, and surface air density ρair. The constants A and B are given by: and θ = 30 is an empirical constant given by Gong [2003]. [17] The source flux correction term suggested by Hoppel et al. [2005] accommodates the disparity between the relatively coarse boundary layer vertical resolution presently afforded in global models and the very fine vertical resolution actually required to model processes in the vertical dimension. It is given by where is the flux into the center of a cell of thickness z, given a source function Sδ appropriate for a flux at height δ (assumed to be 10 m) and the Hoppel correction function fc. The Hoppel correction function is given by where vg is the gravitational settling velocity for the particles of interest, is the midpoint height of our lowest model layer, and Hoppel et al. [2005] suggests the value of unity for β under conditions of neutral stability, while κ = 0.4 is the von Karman constant and u* is the friction velocity in the region of aerosol flux. [18] We have tested this Hoppel-corrected Gong [2003] source function in our model and have compared the model results for surface mass loading to actual sea-salt mass data collected at 11 oceanic locations for the year 1995 (D. L. Savoie and J. M. Prospero, data from the University of Miami network, personal communication, 2004, available from Cooperative Institute for Marine and Atmospheric Studies, Rosenstiel School of Marine and Atmospheric Science, University of Miami, 4600 Rickenbacker Causeway, Miami, Florida 33149–1098, USA, [email protected]). These comparisons are presented as a scatterplot in Figure 2. Figure 2Open in figure viewerPowerPoint The scatterplot comparing our modeled sea-salt aerosol mass using the Hoppel-corrected Gong [2003] sea-salt source function to in situ mass data shows that our model very reasonably reproduces SSA mass loading in the marine surface layer. 6. Data Sets [19] Smirnov et al. [2003] found in an extensive study of AERONET-reported columnar aerosol optical thickness that they were able to clearly separate the contribution of sea salt from other aerosols at Midway Island. Under clean marine conditions, Smirnov et al. [2003] found a linear relationship between 24 h averaged wind speed (〈U10〉24) and marine AOT for mean wind speeds up to 10 m/s (Figure 3a and cases A and B in Table 1). Notably, the data indicate a clear nonzero AOT of 0.018 at zero wind speed, and a most frequently observed AOT of 0.06 at 500 nm in clean marine air masses. Simultaneously, Smirnov et al. [2003] find a negative correlation between wind speed and Ångstrom exponent. The Ångstrom exponent indicates the wavelength dependence of the aerosol extinction coefficient. This negative correlation suggests that at higher wind speeds, aerosol optical thickness at Midway Island is dominated by large (supermicron) particles, presumably sea salt. The data also suggest the presence of very small (submicron) particles, presumably non-sea-salt sulfate (nss), which contribute proportionally more to optical thickness at visible than infrared wavelengths. Figure 3aOpen in figure viewerPowerPoint Clean marine air column optical thickness as predicted by Smirnov et al. [2003] (dashed lines), Mulcahy et al. [2008] (solid lines), and Glantz et al. [2009] (dash-dotted lines) formulations, based on measurements taken at Midway Island (AERONET) and Mace Head, Ireland (PFR), and from SeaWiFS retrievals, respectively. Table 1. Regression Fit Parameters of Sea-Salt Optical Thickness Versus Wind Speed From Models and Observations Case Model τ0 a b r2 Reference Linear Fit of Clean Marine AERONET AOT to 24 h Averaged Wind Speed 〈U10〉24 < 10 m/s A τ500(〈U10〉24) = a · 〈U10〉24 + b – 0.0068 0.056 0.37 Smirnov et al. [2003] B τ1020(〈U10〉24) = a · 〈U10〉24 + b – 0.0093 0.018 0.52 Power Law Fit of Clean Marine Mace Head PFR AOT to Steady State Wind Speed U C τ862(U10) = τ0 + a · U10b 0.035 0.0013 1.95 0.97 Mulcahy et al. [2008] D τ500(U10) = τ0 + a · U10b 0.06 0.00055 2.195 0.97 E τ412(U10) = τ0 + a · U10b 0.032 0.00211 1.755 0.97 F τ368(U10) = τ0 + a · U10b 0.06 0.000356 2.406 0.97 Exponential Fit of Monthly Averaged Marine MODIS AOT to Monthly Averaged Wind Speed U10 G τa,550(U) = τ0ebu f(Λ) – 0.09 0.68 Satheesh et al. [2006] H τSS,550(U) = τ0[ebu − 1] f(Λ) – 0.09 0.68 Power Law Fit of Marine SeaWiFS AOT to Wind Speed U10 I τ555(U) = τ0 + a · Ub 0.036 0.00016 2.3 0.98 Glantz et al. [2009] Linear Fit of Clean Marine AOT to ECMWF Wind Speed U10 J τ550(U10) = a · U10 + b – 0.0040 0.0850 0.95 Huang et al. [2009] Linear Fit of Modeled Sea-Salt AOT to 24 h Averaged Wind Speed 〈U10〉24 < 10 m/s K τ500(〈U10〉24) = a · 〈U10〉24 + b – 0.010 0.020 0.16 this work L τ870(〈U10〉24) = a · 〈U10〉24 + b – 0.010 0.019 0.16 Linear Fit of Modeled Sea-Salt AOT to 24 h Averaged Wind Speed for All 〈U10〉24 M τ870(〈U10〉24) = a · 〈U10〉24 + b – 0.014 0.000 0.29 Linear Fit of Modeled Sea-Salt AOT to Steady State Wind Speed U10 < 10 m/s N τ870(U10) = a · U10 + b – 0.010 0.023 0.12 Linear Fit of Modeled Sea-Salt AOT to Steady State Wind Speed for All U10 O τ870(U10) = a · U10 + b – 0.015 −0.014 0.35 Power Law Fit of Modeled Sea-Salt AOT to 24 h Averaged Wind Speed 〈U10〉24 P τ870(〈U10〉24) = τ0 + a · 〈U10〉24b 0.035 0.00067 2.03 0.19 Q τ870(〈U10〉24) = a · 〈U10〉24b – 0.001 2.05 0.19 Power Law Fit of Modeled Sea-Salt AOT to Steady State U10 R τ870(U10) = τ0 + a · U10b 0.028 0.00073 1.99 0.36 S τ870(U10) = a · U10b – 0.001 2.46 0.28 [20] Using data collected by a Precision Filter Radiometer (PFR) at the GAW station at Mace Head, Ireland, Mulcahy et al. [2008] find good agreement with the results given by Smirnov et al. [2003] for wind speeds up to 10 m/s (Figure 3a and cases C–F in Table 1). Located on the western coast of Ireland, the Mace Head site is well situated to study clean marine air masses moving over Ireland from the North Atlantic. The PFR observes aerosol optical thickness with an estimated precision of 0.01 optical thickness units, centered at λ = 862, 500, 412, and 368 nm. Although breaking coastal waves generate visible aerosol plumes near the shore, relative to an integrated column property such as AOD, they are unlikely to contribute a significant signal. Moreover, Kunz et al. [2002] demonstrated that breaking wave plumes are visible to a lidar, relative to the background marine aerosol backscatter, only at lower winds, and Mulcahy et al. [2008] do not report data points for wind speeds less than 5 m/s. This is significant because sea-salt aerosol production picks up at wind speeds greater than 4 m/s. In other words, at higher wind speeds, the background breaking oceanic waves generate sea-salt aerosol fields with optical thickness greater than the optical thickness of coastal surf plumes. Mulcahy et al. [2008] find a relationship of the form τ(U) = τ0 + a · Ub between marine aerosol optical thickness and surface wind speed for wind speeds up to U = 18 m s−1, a higher wind speed than was included in the observations of Smirnov et al. [2003]. [21] Instead of averaging wind speed over the previous 24 h period, as Smirnov et al. [2003] did, to determine the relationship between clean marine optical thickness and wind speed, Mulcahy et al. [2008] filter their data to include only optical thickness measurements taken under steady state wind speed conditions. Steady state wind conditions require that the maximum standard deviation in wind speed not exceed 2 m s−1 for the entire day on which data was taken, with a standard deviation in wind speed of less than 1 m s−1 during the two minute interval over which the PFR recorded optical thickness. These parameters were relaxed to 3 m s−1 and 2 m s−1 for days with average wind speeds greater than 10 m s−1. Additionally, Mulcahy et al. [2008] specify conditions of maximum black carbon mass concentration of 50 ng m−3, total particle number concentration less than 700 cm−3, and origin of a given air mass from the clean marine sector west of Mace Head. These requirements substantially decrease the number of data points available for performing a regression analysis of AOT against wind speed, with only 14 days out of 10 months (November–March for 2 ye