DMS gas transfer coefficients from algal blooms in the Southern Ocean

Introduction Conclusions References


Introduction
Gas exchange across the ocean-atmosphere interface influences the atmospheric abundance of many compounds of importance to climate and air quality.Such compounds include greenhouse gases, aerosol precursors, stratospheric ozone-depleting substances, and a wide range of photochemically reactive volatile organic carbon compounds that influence tropospheric ozone.Estimating the air/sea fluxes of all of these Introduction

Conclusions References
Tables Figures

Back Close
Full compounds requires knowledge of their distributions in near surface air and seawater and an understanding of the transport processes controlling gas exchange across the air/sea interface.The transport processes are not well understood, in large part because of the paucity of direct air/sea gas flux observations.The parameterization of gas exchange is a significant source of uncertainty in ocean/atmosphere exchange in global models, particularly at high wind speeds (Elliott, 2009).Gas flux is typically calculated using the concentration gradient across the air/sea interface (∆C) and the gas transfer coefficient (K ): K represents the inverse of the resistance to gas transfer on both the water and air sides of the interface (i.e.1/K = r w + r a ) and can be expressed in either waterside or airside units (Liss and Slater, 1974).Equation ( 1) is a very simple expression that belies the complex physical process involving diffusive and turbulent mixing at the boundary between two mediums of very different densities.Wind stress is the predominant forcing for gas transfer, but mixing at the interface is also influenced by buoyancy, windwave interactions, wave breaking, surfactants, and bubble generation.The interface is chemically complex owing to the presence of organic films or particles, and for some gases the interface may be biologically/photochemically reactive.Most air/sea gas transfer calculations utilize wind speed-based parameterizations derived from deliberate dual tracer observations (Ho et al., 2011;Nightingale et al., 2000), sometimes scaled to agree with the long-term global average oceanic uptake of 14 CO 2 (Sweeney et al., 2007).The dual tracer technique is a waterside method that requires data averaging over periods of hours to days, thus averaging over significant changes in conditions.Eddy covariance is a direct flux measurement carried out on the air side of the interface.In conjunction with measurements of the air/sea concentration difference, eddy covariance studies can determine the gas transfer coefficient, K , on short time scales (10 min-1 h).This provides a capability to assess variability in K due to the influence of rapid changes in near surface processes (e.g.wind-wave interac-Introduction

Conclusions References
Tables Figures

Back Close
Full tions, bubbles, surfactants).Eddy covariance requires high frequency sensors, and flux studies to date have been carried out on only a few compounds (Huebert et al., 2004;McGillis et al., 2001;Yang et al., 2013;Kim et al., 2014;Blomquist et al., 2012;Bariteau et al., 2010;Marandino et al., 2005).DMS air/sea transfer resistance is predominantly on the water side, a characteristic it shares with CO 2 .DMS is moderately soluble and weakly influenced by bubblemediated gas transfer, in contrast to CO 2 , which is sparingly soluble and strongly influenced by bubble-mediated gas transfer.This makes DMS a useful tracer for watersidecontrolled, interfacial gas transfer.Measurements of gas exchange using insoluble gases have suggested that the relationship between K and wind speed is non-linear (Nightingale et al., 2000;Sweeney et al., 2007;Miller et al., 2010;Ho et al., 2011).
In contrast, the majority of DMS eddy covariance data suggests a linear relationship between K and wind speed (Yang et al., 2011).Blomquist et al. (2006) suggest that the differences in functional form of these relationships may be due to the disproportionate influence of bubbles upon the flux of insoluble gases (Woolf, 1997).
Physical process models have made significant progress in parameterizing gas exchange with input terms that include but are not limited to wind speed.However, these models are still in development and are capable of substantially different estimates of K depending on how non-wind speed terms such as wind-wave dynamics are applied in the model (Fairall et al., 2011;Soloviev, 2007).Bell et al. (2013) recently demonstrated that some of the scatter in eddy covariance measurements may be explained by spatial/temporal differences in wind-wave interaction, although the role of surfactants cannot be ruled out.Gas exchange measurements in an artificial surfactant patch (Salter et al., 2011) and in laboratory studies using natural surfactants (Frew et al., 1990) have demonstrated marked suppression of gas transfer.Additional eddy covariance gas exchange observations are required to improve these gas exchange models.Eddy covariance DMS flux measurements have been made in the Atlantic Ocean (Bell et al., 2013;Marandino et al., 2008;Salter et al., 2011;Blomquist et al., 2006) and Pacific Ocean (Marandino et al., 2007(Marandino et al., , 2009;;Yang et al., 2009), with three of these studies at Introduction

Conclusions References
Tables Figures

Back Close
Full high northern latitudes.Only one previous study has been performed in the Southern Ocean (Yang et al., 2011).
The Southern Ocean has a unique wind and wave environment: minimal land mass in the Southern Hemisphere leads to strong, consistent winds and waves with a long fetch.The duration of the wind speed event rather than the wind fetch is the most important factor influencing the waves (Smith et al., 2011).This region is very important in determining the global uptake of atmospheric CO 2 by the ocean (Sabine et al., 2004) and the supply of DMS as a source of atmospheric sulfate aerosol (Lana et al., 2011).This paper presents data collected in the Southern Ocean summer (February-March 2012) as part of the New Zealand Surface Ocean Aerosol Production (SOAP) cruise (Fig. 1).During the cruise, a variety of oceanic, atmospheric and flux measurements were collected.The cruise targeted regions of extremely high biological activity (blooms of dinoflagellates and coccolithophores) and encountered a number of atmospheric frontal events leading to winds in excess of 11 m s −1 .

Mast-mounted instrumentation and data acquisition setup
The eddy covariance setup was mounted on the bow mast of the R/V Tangaroa, 12.6 m above the sea surface.Three dimensional winds and sonic temperature (Campbell CSAT3) and platform angular rates and accelerations (Systron Donner Motion Pak II) were measured on the mast and co-located with the air sampling inlets for DMS.Air was drawn through the sampling inlets at 90 SLPM under fully turbulent flow conditions (Re > 10 000).Analog signals from all of these instruments were filtered at 15 Hz and then logged at 50 Hz (National Instruments SCXI-1143).The ship's compass and GPS systems were digitally logged at 1 Hz.The mast configuration was similar to that used during the Knorr_11 North Atlantic cruise (Bell et al., 2013), with the following two changes: Introduction

Conclusions References
Tables Figures

Back Close
Full 1.An air sampling inlet with integral ports for standard delivery was fabricated from a solid block of PTFE.The design minimized regions of dead space that might attenuate high frequency fluctuations and result in loss of flux signal.
2. A shorter length of 3/8 ID Teflon tubing was used between the mast and the container van.A 19 m inlet was used during SOAP in contrast to the 28 m inlet used during Knorr_11 (Bell et al., 2013).

Atmospheric and seawater DMS
DMS was measured in air and in gas equilibrated with seawater using two atmospheric pressure chemical ionization mass spectrometers (Bell et al., 2013).In both instruments, a heated (400 • C) radioactive nickel foil (Ni-63) generates protons that associate with water molecule clusters in the sample stream.Protonated water vapor (H 3 O + ) undergoes a charge transfer reaction to form protonated DMS ions (m/z = 63) that are then quadrupole mass filtered and counted.Tri-deuterated DMS (d3-DMS, m/z = 66) was used as an internal standard for both instruments.
Atmospheric measurements were made with the University of California, Irvine (UCI) mesoCIMS instrument (Bell et al., 2013).A gaseous d3-DMS standard was introduced to the atmospheric sample stream at the air inlet via a 3-way valve mounted at the base of the bow mast.The gas standard was diverted to waste every 4 h and the response of the d3-DMS signal recorded as a measure of the inlet tubing impact on signal delay and frequency loss.Air from the bow mast was sub-sampled at approximately 1 L min −1 and DMS levels were calculated as follows: Where S 63 and S 66 represent blank-corrected signals from DMS and d3-DMS respectively (Hz), F Std and F Total are the gas flow rates of the d3-DMS standard and the inlet air (L min −1 ), and C Tank is the gas standard mixing ratio.Introduction

Conclusions References
Tables Figures

Back Close
Full Seawater measurements were made with a smaller instrument (UCI miniCIMS), which utilizes a modified residual gas analyzer as the mass filter and ion detector (Stanford Research Systems RGA-200;Saltzman et al., 2009).Aqueous d3-DMS standard was delivered by a syringe pump (New-Era NE300) to the ship's underway seawater supply upstream of the equilibrator (see Bell et al., 2013, for details).The natural DMS and the d3-DMS standard are both transported across the membrane and the DMS concentration in seawater in the equilibrator is then calculated as follows: Sig 63 and Sig 66 represent the average blank-corrected ion currents (pA) of protonated DMS (m/z = 63) and d3-DMS (m/z = 66), respectively, C Std is the concentration of d3-DMS liquid standard (nM), F Syr is the syringe pump flow rate (L min −1 ), and F sw is the seawater flow rate (L min −1 ).Seawater concentrations were averaged at 1 min intervals for the entire SOAP dataset.Lag correlation analysis between the ship surface seawater temperature and equilibrator temperature records identified that a 3-4 min adjustment in the DMS sw was required to account for the delay between water entering the seawater intake beneath the hull of the ship and it reaching the miniCIMS equilibrator.
We compared our seawater measurements with discrete samples collected by the NIWA team and analysed using sulfur chemiluminescence detection (SCD).The NIWA discrete analyses were performed on water collected from both the underway supply and from CTD Niskin bottles fired in the near surface (< 10 m).The analytical techniques (SCD and miniCIMS) typically agreed well and these results will be discussed elsewhere.Throughout the cruise, data from the underway and CTD bottles were in good agreement (Fig. 2), with the exception of Day of Year (DOY) 54-55 when the ship's underway supply became significantly contaminated.The contamination was biological and resulted in DMS levels at least twofold higher than from a Niskin bottle fired at the same depth.Flushing and soaking the underway lines in a biologically-Introduction

Conclusions References
Tables Figures

Back Close
Full active cleaning solution (Gamazyme™) and cleaning the equilibrator with dilute (10 %) hydrochloric acid resolved the problem.The data from DOY 54-55 has been excluded from our analysis.

DMS flux calculation: eddy covariance data processing and quality control
Air/sea flux calculation involved the same procedure detailed in Bell et al. (2013).
Apparent winds were corrected for ship motion according to the procedures of Edson et al. (1998) and Miller et al. (2008).Ten minute flux intervals with a mean relative wind direction within ±90 • (where winds onto the bow = 0 • ) were retained for subsequent data analysis.The DMS signal was adjusted relative to the wind signals to account for the timing delay due to the inlet tubing.The delay was estimated to be 1.9 s from the periodic firing of a 3-way valve on the bow mast.An equivalent delay estimate was ascertained by optimization of the cross correlation between DMS and vertical wind.Flux intervals were computed from the co-variation in fluctuations in vertical winds (w ) and DMS (c ) flux.The internal d3-DMS standard exhibited negligible covariance with vertical wind, confirming that no density correction due to water vapor or temperature fluctuations (i.e."Webb" correction) was required for our DMS fluxes.
Cospectral analysis objectively removed intervals with large low frequency fluctuations and the criteria for elimination is defined in Bell et al. (2013).This process reduced scatter in the data without introducing an obvious bias.High frequency flux loss in the inlet tubing was estimated by modeling a filter based on the d3-DMS signal attenuation when the bow mast valve was switched.The inverse filter was then applied to wind speed binned DMS cospectra.This enabled an estimate of the necessary wind speed-dependent high frequency loss correction (Flux Gain = 0.004U 10n + 1.012).Introduction

Conclusions References
Tables Figures

Back Close
Full

DMS gas transfer velocity calculation
Gas transfer velocities were calculated following: Where F DMS is the measured DMS air/sea flux (mol m −2 s −1 ), DMS sw is the seawater DMS level (mol m −3 ), DMS air is the atmospheric DMS partial pressure (atm), and H DMS is the temperature-dependent DMS solubility in seawater (atm m 3 mol −1 ; Dacey et al., 1984).K DMS values were calculated from the cruise data using 10 min averages.
The water side only gas transfer coefficient, k w , was obtained from the expression: Where K DMS is the total DMS gas transfer coefficient, α is the dimensionless Henry's Law constant for DMS, and k a is the air side gas transfer coefficient.In situ k a values were obtained from NOAA COARE driven by in situ measurements of wind speed, atmospheric pressure, humidity, irradiance and air and seawater temperature.The average (mean) difference between k w and K DMS was 7 %.In order to compare our results with various other gas transfer parameterizations, k w was then normalized to a Schmidt number of 660 (CO 2 at 25 • C): Where Sc DMS is calculated using the ship's seawater temperature recorded at the bow and Eqn. 15 in Saltzman et al. (1993).Introduction

Conclusions References
Tables Figures

Back Close
Full

Cruise track, meteorological, and oceanographic setting
The SOAP cruise sampling strategy was to identify phytoplankton blooms using ocean color imagery and then use underway sensors (e.g.chlorophyll a fluorescence, DMS) to map out the in situ spatial distribution.Three blooms were identified and sampled: B1, B2 and B3 (Fig. 1).B1 was an intense dinoflagellate-dominated bloom at approx.44.5 and fluorescence (0.99 ± 0.35 mg m −3 ).After sampling B2, the B1 location was revisited and a new bloom (B3) was identified with a mixed population of coccolithophores, flagellates and dinoflagellates (DOY 57.9-60.5).B3 DMS levels (5.9 ± 1.5 nM) were substantially lower than in B1.
The time series plot in Fig. 2 describes the oceanographic and meteorological variability throughout the cruise.Surface ocean temperatures (SSTs) were consistent at 14.7 ± 1.0 • C while atmospheric temperature fluctuated just above and below the SST.

Wind speed dependence of gas transfer coefficients
The SOAP gas transfer coefficients exhibit a positive correlation with wind speed (Spearman's ρ = 0.57, p < 0.01, n = 1327; Fig. 3, left panel).A linear least squares fit to the data gives k 660 = 2.17 ± 0.10U 10n -0.51 ± 0.91 with an adjusted R 2 = 0.25.As with previous shipboard eddy covariance DMS studies, using a second order polynomial does not improve the fit to the data (adjusted R 2 = 0.25).The linear model is not well suited for this data set because the residuals are not normally distributed (Fig. 3, right panel).The frequency distribution of the SOAP k 660 measurements exhibits positive skewness at all wind speeds (Fig. B, Supplement).The skew in the SOAP k 660 data appears to originate in the frequency distribution of seawater DMS.Surface ocean DMS distributions are typically characterized by positive skew and this is evident in the global surface ocean DMS database (Lana et al., 2011).It is not surprising to see skewed distributions in the SOAP data as the cruise encountered strong, non-linear gradients in biological activity.There is no skewness in the distribution of winds within each wind speed bin.Skewness in the seawater DMS distribution should propagate into the DMS flux distribution simply because air/sea flux is proportional to air/sea concentration gradient, which is controlled in turn by seawater DMS levels (Figs. C and D, Supplement).If F DMS and ∆C are highly correlated, then the variance in k 660 should be considerably less than that in either parameter and would exhibit less skew.This is not the case: k 660 exhibits a similar skew to F DMS and ∆C.As a result of the frequency distribution observations in the SOAP dataset, we reexamined data from a recent North Atlantic cruise (Bell et al., 2013;Figs. E-G, Supplement).The frequency distributions of k 660 , F DMS and DMS sw exhibit similar positive skewness to that in the SOAP data set.In order to better represent the central tendency of the k 660 data and assess the relationship with wind speed, geometric means were computed for 1 m s −1 wind speed bins (Fig. 4).Binned k 660 data from both cruises demonstrate a shallower slope using the geometric means.
The SOAP k 660 bin average data (Fig. 5) exhibit a linear relationship with wind speed for low and intermediate winds, as found in previous DMS flux studies (e.g.Huebert et al., 2010;Yang et al., 2011;Marandino et al., 2007Marandino et al., , 2009)).For wind speeds up to 14 m s −1 , the binned geometric mean SOAP data yields a linear regression equation of k 660 = 2.07U 10n -2.42, which is slightly shallower than that obtained from a compilation of previously published DMS gas transfer measurements (k 660 = 2.6U 10n -5.7; Goddijn-Murphy et al., 2012).In the higher wind speed bins (above 10 m s −1 ), the relationship between k 660 and wind appears to weaken.A weaker relationship between k 660 and wind speed at high wind speeds was also observed in the North Atlantic (Bell et al., 2013).In both cruises, there is limited data at wind speeds above 10 m s −1 , so this phenomenon should be viewed with caution.Bell et al. (2013) suggested that the effect could be due to suppression of near surface turbulence due to wind/wave interactions (Soloviev et al., 2007;Donelan et al., 2010).
The SOAP study did not include direct measurements of wave properties or surfactants.Significant wave height was estimated using satellite reanalysis products from ECMWF and NCEP, which agreed well (Spearman's ρ = 0.91, p < 0.01, n = 2876).Significant wave height exceeded 4.5 m during SOAP.There is no obvious relationship between significant wave height and the scatter in the relationship between gas transfer and horizontal wind speed during SOAP (Fig. H, Supplement).In situ fluorescence was used as an indicator of biological activity during SOAP.Fluorescence sensors were located in seawater continuously pumped through the ship from the near surface intake beneath the hull.The variability in the gas transfer velocity data is not explained by Introduction

Conclusions References
Tables Figures

Back Close
Full  Supplement).Note that fluorescence is not necessarily a reliable indicator of surfactant concentrations.The relative importance of waves and/or surfactants in air/sea gas exchange remains unclear and requires dedicated measurements to be made concurrent with direct assessments of gas exchange by eddy covariance.

Uncertainties in K introduced by flux footprint and seawater DMS heterogeneity
As discussed above, spatial heterogeneity of seawater DMS can introduce uncertainty in gas transfer coefficients derived from eddy covariance studies.It is logistically challenging to quantify footprint effects from a single ship, and it has not been done on prior studies.On the SOAP cruise, the fortuitous alignment of winds and ship track downwind of the dinoflagellate-dominated bloom (B1) provided a unique opportunity to quantify the length scale associated with the flux footprint.
The SOAP cruise spent approximately 5 days mapping out the spatial extent of B1 waters, then transited out of the bloom to WP1 about 150 km to the southwest.The ship then steamed back into and across B1 at a ship speed of 5.1 ± 0.7 m s −1 , over about 18 h (DOY 50.85-51.35;Fig. 6).Meteorological and oceanographic conditions were relatively constant during the B1 transect, with wind speeds ranging from 5.5-9.7 m s −1 , wind direction from 5-33 • , air temperature of 15.4±0.8, and SST of 14.4±0.5 (Fig. 7).Atmospheric stability was neutral-stable during this period.A detailed picture of surface ocean DMS levels in and around B1 can be seen from the data collected between DOY 45.65 and DOY 51.35 (Fig. 6).DMS levels exhibit a sharp step-change at approximately 44.6 • S. DMS concentrations south of the bloom were less than 5 nM.
Near the bloom center, levels increased rapidly over a few kilometers from below 10 nM to greater than 15 nM.Atmospheric DMS levels were quite stable during the transect with a mean of 489 ± 58 ppt.The ship's heading (approx.27 • ) meant that winds blew almost directly onto the bow, with < 10 • difference for the final 60 km of the transect back into B1.28466 Introduction

Conclusions References
Tables Figures

Back Close
Full Figure 7 depicts seawater DMS levels (green symbols) as the ship steamed into B1 waters.DMS levels 120 km away from the bloom were below 5 nM and consistently 5-10 nM until the southern perimeter of the bloom (0 km).DMS levels increased rapidly to 15-20 nM as the ship moved into the bloom.DMS flux divided by the horizontal wind speed is also presented.We assume a relatively linear relationship between k 660 and U 10n and that fluctuations in F DMS /U 10n (Fig. 7, blue symbols) are driven primarily by changes in ∆C (i.e.DMS sw ).Spikes in F DMS /U 10n are evident in DMS sw after a consistent distance/time lag. Figure 7 also plots gas transfer velocities (lower panel) during the transect into B1.COARE model output for DMS is plotted as a reference line.Spikes in k 660 are coincident with sharp changes in F DMS /U 10n prior to the lagged corresponding change in DMS sw .
On this transect the eddy covariance flux footprint was directly ahead of the ship, so a lag would be expected between the F DMS and DMS sw .The maximum correlation between F DMS /U 10n (using the midpoint of the flux interval) and DMS sw was obtained for a lag of 8 min.This lag represents a distance of ∼ 2.5 km at 5.1 m s −1 ship speed.Applying this lag to the calculation of gas transfer velocity reduced the scatter (Fig. 8).We compared the flux footprint obtained from the lag calculation to a flux footprint calculation using an online version of an analytic dispersion model (http://www.geos.ed.ac.uk/abs/research/micromet/java/flux.html;Kormann and Meixner, 2001).We ran the model with representative conditions for the SOAP B1 tran- tribution range from 0.3 to 1.9 km.Model runs where measurement height was varied to reflect the limits of ship motion (significant wave height from ECMWF suggests the vertical displacement of the flux inlet was at least 2.5 m) gave minimum and maximum peak flux footprint contributions of 0.4 and 2.0 km respectively.Despite the sensitivity of the model to the input parameters, none of these estimates are as large as the footprint derived from the lag calculation.Huebert et al. (2010) addressed surface ocean spatial heterogeneity for their estimates of DMS gas transfer velocity during the June 2007 Deep Ocean Gas Exchange Experiment (DOGEE) in the North Atlantic.When hourly DMS sw relative standard error of the mean (RSEM) exceeded 0.25, gas exchange data was not included in their analysis.Removing k 660 data with high DMS sw variability during DOGEE improved the correlation between k 660 and wind speed.We assessed variability in our high frequency DMS sw data by calculating the forward-looking, running SD (SD) on a one hour timescale.The relative SD (RSD) was then calculated by dividing the SD by DMS sw .Using the RSD would not have been reliable for identifying the outlying k 660 data during the B1 transect (Fig. 8, left panel).The scatter in k 660 vs. U 10n in the entire SOAP dataset cannot be reduced on the basis of the associated RSD values (Fig. J, Supplement).

Conclusions
The SOAP k 660 bin average values are in good agreement with previous gas transfer studies using eddy covariance of DMS (Yang et al., 2011;Bell et al., 2013;Marandino et al., 2007).As noted earlier, these studies provide evidence that interfacial gas transfer is a relatively linear function of wind speed for low-intermediate wind speeds.There is some evidence that the dependence on wind speed weakens at higher wind speeds both in this study and in the Knorr_11 study (Bell et al., 2013).There is no evidence in any of the DMS eddy covariance data sets that the interfacial (non-bubble mediated) component of gas transfer has a wind speed-dependence greater than linear.How-Introduction

Conclusions References
Tables Figures

Back Close
Full ever there is still very limited data above 10 m s −1 and the high wind speed trends are uncertain.
The scatter in the SOAP data is typical of shipboard eddy covariance flux measurements.This arises from fluctuations in near surface turbulence and vertical entrainment, vertical shear, ship motion, heterogeneity in seawater DMS and variations in atmospheric DMS due to chemical losses (Blomquist et al., 2010).We note the skewness of the gas transfer velocities in a given wind speed range and use geometric statistics to characterize the central tendency and variance of the data.This skewness is likely driven by the inherent lognormal distribution of seawater DMS levels.We propose that spatial heterogeneity in seawater DMS causes decorrelation between the measured seawater DMS and the observed DMS flux, which results in skewness propagating into the calculated transfer coefficients.The data from this study may be particularly influenced by the large differences in seawater DMS values inside and outside the phytoplankton blooms.Similar skewness was observed in data from the North Atlantic ocean (Bell et al., 2013) and this phenomenon likely affects all DMS eddy covariance studies to some degree.If so, then some transformation of the DMS gas transfer velocities is warranted.
The transect from WP1 into B1 provided a unique opportunity to quantitatively estimate the spatial extent of the eddy covariance flux footprint.The data suggest that the shipboard flux measurements were sensitive to changes in seawater DMS approximately 2.5 km upwind of the ship, a surprisingly large distance.This transect was conducted under neutral to stable conditions, when one might expect the flux footprint to be relatively large.This result is much greater (twofold or more) than that predicted using an analytic dispersion model (Kormann and Meixner, 2001).A flux footprint model developed for marine air/sea gas flux measurements would be an invaluable tool for the ocean/atmosphere gas exchange research community.
During the SOAP cruise we saw no obvious evidence of a first order biological effect on gas transfer coefficients.From this it could be inferred that surfactants in the dinoflagellate and coccolithophore blooms did not exert a significant effect on water Introduction

Conclusions References
Tables Figures

Back Close
Full side turbulence.Any modification of the gas transfer velocity vs. wind speed relationship by surfactants or waves during SOAP was masked by other influences upon the variability in gas flux measurements.Minimizing the scatter in gas transfer velocity is critical in order to observe the influence of non-wind speed processes and to draw firm conclusions about their impact upon air/sea gas transfer.The challenge for the gas exchange community is that heterogeneity in seawater DMS concentrations is linked to phytoplankton growth, which likely also determines surfactant effects upon the gas transfer velocity.
Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | 3 Results Weather systems from the North brought relatively warm air and systems from the South brought cooler air.For example, the atmospheric front on DOY 55 from the South caused air temperatures to drop from approximately 18 to 12 • C (Fig.2, upper panel).Frontal systems passed over the ship regularly throughout the cruise and the final system (DOY 61.6-64) brought intense winds from the North.During SOAP, the horizontal wind speeds predominantly ranged between 1 and 15 m s −1 .The atmospheric boundary layer was stable (z/L > 0.05) for approximately 25 % of the cruise (Fig.2, upper panel).Yang et al. (2011) suggest that a stable boundary layer leads to greater scatter 28462 Discussion Paper | Discussion Paper | Discussion Paper | and a potentially negative bias in k 660 vs. wind speed plots.Our data do not suggest increased scatter or any bias during stable periods (Fig. A, Supplement) and we have not filtered the SOAP k 660 data on this basis.Oceanic and atmospheric DMS levels were extremely high during the first half of the cruise (DOY 44-54; Fig. 2, middle panel).The majority of this period was spent in and around B1 waters, with elevated seawater DMS (> 10 nM) and atmospheric DMS (> 600 ppt).Oceanic DMS was always at least an order of magnitude greater than atmospheric DMS, meaning that the air/sea concentration gradient was effectively controlled by DMS sw .The second half of the cruise (DOY 55-65) encountered less productive blooms with lower seawater DMS levels.The reduction in oceanic DMS was mirrored by lower atmospheric DMS levels (151 ± 73 ppt, DOY 55-65).Ten minute average DMS fluxes (F DMS ) measured by eddy covariance are plotted in Fig. 2. F DMS reflected the seawater DMS levels, with three notable peaks while inside B1 waters (> 60 µmol m −2 day −1 , DOY 48-50).F DMS was generally lower during the second half of the cruise (13±10 µmol m −2 day −1 , DOY 55-65) but elevated fluxes were still observed due to increased horizontal wind speeds (e.g.approx.45 µmol m −2 day −1 on DOY 61.6).SOAP gas transfer coefficients were calculated at 10 min intervals (Fig. 2, lower panel) following Eqs.(1)-(6) using measurements of F DMS , oceanic and atmospheric DMS levels and SST.During some periods of constant wind speed, the NOAA COARE (v3.1) estimates are close to the observed k 660 values (e.g.DOY 51).However, at various times during the cruise, the NOAA COARE estimates exhibit significant divergence from the observed k 660 values.The difference was sometimes positive, as on DOY 48 and sometimes negative, as on DOY 53.These divergences are not random scatter about the COARE prediction and suggest that unaccounted-for processes are influencing our measurements of gas transfer.Discussion Paper | Discussion Paper | Discussion Paper | For example, the correlation coefficient between DMS flux and seawater concentration in the 13-14 m s −1 wind speed bin (Spearman's ρ = 0.45, p < 0.01, n = 47) is considerably lower than expected.Decorrelation of DMS flux and seawater concentration is likely due to mismatches between seawater DMS levels measured aboard ship and those in the actual footprint of the flux.Misalignment between seawater DMS levels and the flux footprint is virtually unavoidable in a region of strong spatial heterogeneity, where wind direction and ship track are never perfectly aligned.Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | surface ocean fluorescence (Fig. I, Discussion Paper | Discussion Paper | Discussion Paper | sect: measurement height = 12 m; wind speed = 8 m s −1 ; roughness length = 0.02 m (minimum value available); zero-plane displacement = 0.5 m (minimum value available); sensible heat flux = −20 W m −2 ; air temperature = 15 • C. The footprint model predicts a peak relative flux contribution (defined as 90 % of the relative flux) 0.8 km ahead of the ship, less than half of the distance inferred from the field observations.The calculated footprint is highly sensitive to the input parameters.During the SOAP B1 transect, atmospheric stability was slightly stable but close to neutral (z L −1 ∼ +0.1).Relatively small changes in wind speed (±1 m s −1 ), temperature (±1 • C) or sensible heat flux (+10 W m −2 ) alter the stability such that model predictions of the peak footprint con-Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper |

Figure 1 .
Figure 1.Cruise track during the SOAP study, which began and finished in Wellington, New Zealand.The phytoplankton blooms (B1-3) and waypoint 1 (WP1) locations are identified.

Figure 2 .
Figure 2. Time series data (10 min averages) from the SOAP cruise.Dashed black line on top panel indicates neutral atmospheric stability (z/L = 0).SOAP k 660 data are divided into on station (squares, ship speed < 1.5 m s −1 ) and off station (circles, ship speed ≥ 1.5 m s −1 ).