Strong relationship between dimethyl sulfide and net community production in the western subarctic Pacific

Authors


Abstract

[1] Although much attention has been paid to describing the distribution of oceanic dimethyl sulfide (DMS) concentrations, establishing robust relationships between DMS concentrations and biological, physical, and chemical variables is still challenging. Previous studies have proposed semiempirical parameterizations by combining multiple physical and biogeochemical parameters to better understand and reproduce the global distribution of sea surface DMS. However, none of these parameterization schemes could reconcile regionally elevated DMS peaks found in high-resolution DMS measurements made in the western subarctic Pacific. Here we found that DMS concentrations are highly correlated with the net community production, a parameter that integrates biological activity over time. We anticipate that this relationship may be exportable to other regions with high primary productivity, such as the Southern Ocean or upwelling regions, and can be used as an important parameterization scheme, combined with solar radiation dose relationship.

1 Introduction

[2] Dimethyl sulfide (DMS) plays important roles in the Earth' climate system and the global biogeochemical cycles, with the oceanic emission being the main natural source of DMS into the atmosphere. In the atmosphere, DMS is photooxidized to form sulfate aerosols, affecting the radiative budget of the atmosphere by serving as precursors of cloud condensation nuclei (CCN). A climate feedback role of the DMS production has been referred to as the Charlson-Lovelock-Andreae-Warren (CLAW) hypothesis [Charlson et al., 1987]. Although accumulated evidence since then suggests that marine sources of CCN to the overlying atmosphere and the response of clouds to changes in aerosol are much more complicated than was recognized when the CLAW hypothesis was suggested [Quinn and Bates, 2011], biogenic sulfur emissions from the Southern Ocean, where anthropogenic aerosol sources are negligible, have been greatly underestimated and could be further enhanced owing to climate warming [Cameron-Smith et al., 2011; Levasseur, 2011]. This highlights importance of oceanic DMS emissions in regulating the Earth' climate in the regional scale.

[3] The production of DMS in the ocean occurs exclusively through biogenic processes [Kettle et al., 1999]. The fact that the production pathways are complicated [Stefels et al., 2007], however, still prevents a comprehensive revelation of the important processes and factors that regulate DMS distribution in the ocean. Because DMS production is closely associated with biological activities, its distribution is also regulated by in situ biodiversity, especially in regions with high biological productivity. Consequently, it is important and necessary to better understand the factors that control DMS production and consumption processes and thereby to predict corresponding changes in the distribution of DMS, with subsequent changes in CCN production and hence the efficiency of solar radiation reflection.

[4] Global climatologies have been reported based on data sets of observed DMS concentrations and model-reconstructed fields of surface DMS concentrations and fluxes to the atmosphere [Kettle et al., 1999; Lana et al., 2011]. Thus, much attention has been paid to describing DMS concentration in the ocean during the past decade [Simó and Dachs, 2002; Vallina and Simó, 2007; Watanabe et al., 2007; Miles et al., 2012]. The solar radiation dose (SRD) is generally believed to be one of the important factors that regulate DMS distributions in the sea surface [Vallina and Simó, 2007], and the relationship between SRDs and DMS concentrations is widely used in climate models to infer the impacts of DMS on climate change [Le Clainche et al., 2010]. However, global models that incorporate the relationship between DMS and SRD do not fully capture DMS variability, and therefore, factors other than SRD are important in determining global DMS emissions [Derevianko et al., 2009].

[5] Here we reexamined several proposed algorithms and used them to reconstruct sea surface DMS concentrations that we compared to a highly time-resolved DMS data set from the western North Pacific Ocean, where primary production rates are among the highest in the world oceans [Longhurst et al., 1995] and DMS data are lacking relative to the total sea surface area [Lana et al., 2011]. We focused on this region to test and improve existing algorithms. After comparing the correlations between DMS concentrations and biogeochemical parameters, we propose a new scheme based on net community production (NCP) for reconstructing sea surface DMS concentrations in the area. Finally, we discuss the importance of using an appropriate climatological parameterization for each biogeochemical setting in calculating sea surface DMS concentrations and estimating DMS fluxes between the ocean and atmosphere.

2 EI-PTR-MS Instrumentation and 2008 Subarctic Pacific Experiment for Ecosystem Dynamics Study (SPEEDS)/Surface Ocean Lower Atmosphere Study (SOLAS) Cruise

[6] The detailed setup of the equilibrator inlet-proton transfer reaction-mass spectrometry (EI-PTR-MS) system has been described previously [Kameyama et al., 2009]. Briefly, the EI-PTR-MS system consists of a glass bubbling-type equilibrator and a commercially available PTR-MS instrument (Ionicon Analytik GmbH, Innsbruck, Austria). Seawater was continuously introduced to the equilibrator at 1.0 L min−1. Dissolved DMS in the sample was extracted into a DMS-free carrier N2 gas stream at complete equilibrium between liquid and gas phases within the equilibrator, and part of the carrier gas including extracted DMS was continuously directed to the PTR-MS without pretreatment. The presence of DMS was indicated by ion signals at m/z = 63 (CH3SCH3·H+).

[7] Field observations were made during the 2008 SPEEDS/SOLAS cruise of R/V Hakuho Maru in the western subarctic North Pacific Ocean in July and August 2008 (see Figure S1 in the supporting information). The source of seawater for EI-PTR-MS measurements was either the ship' seawater sampling system, which pumped seawater from an intake on the keel at a depth of about 5 m, or a submersible torpedo-shaped fish equipped with a clean pump system suspended from the ship while underway. During the cruise, continuous measurements of biochemical parameters including sea surface temperature (SST), sea surface salinity (SSS), sea surface nitrate (SSN), chlorophyll a, and pCO2 in surface seawater were carried out. NCP was calculated as the difference in salinity-normalized dissolved inorganic carbon (NDIC) between winter and the time of observation in summer (NDICobs). NDICobs at S = 33 was computed from pCO2, SST, SSS, and salinity-normalized total alkalinity. The details of the measurements and calculation are shown in the supporting information.

3 Results and Discussion

[8] Figure 1 shows the time series of DMS concentrations and physical and biogeochemical parameters at the sea surface, obtained north of 44°N in the western Pacific. This region is considered part of the Pacific SubArctic gyres (West) (PSAW), of the biogeographic provinces [Longhurst, 1998]. Small variations in salinity and temperature in the region indicate that the water column in this region is generally physicochemically stable. There were large variations and several elevated peaks in the concentrations of DMS observed during the cruise (Figure 1c). These peaks were observed at spatial scales less than several kilometers (i.e., at temporal scale less than several hours). There were good correlations between DMS and chlorophyll a concentrations at low DMS levels (<5 nmol L−1, r2 = 0.43, P < 0.05, n = 333, Spearman two-tailed test). However, the regional DMS peaks (6–10 nmol L−1), or “hot spots,” were not always associated with phytoplankton blooms (e.g., on 5–7 August 2008). Interestingly, the variation of net community production paralleled that of DMS concentration better than chlorophyll a over the entire DMS concentration range, and particularly at the peak levels. NCP is equivalent to gross primary production minus respiration by all autotrophic and heterotrophic organisms. NCP integrated over the period from the preceding late winter to the time of observation is calculated as the change in NDIC for that period, corrected for the production of mineral calcium carbonate and the net air-sea CO2 flux [Ishii et al., 1998].

Figure 1.

Time series of (a) latitude and longitude, (b) sea surface temperature and salinity, and (c) DMS concentrations (red dots), NCP (blue dots), and chlorophyll a concentrations (green lines) during the 2008 SPEEDS/SOLAS cruise of R/V Hakuho Maru in the western North Pacific Ocean.

[9] Biochemical parameters were used to reconstruct DMS concentrations at the ocean surface as Simó and Dachs [2002],

display math(1)
display math(2)

and as Watanabe et al. [2007],

display math(3)

where MLD is the mixed layer depth (m), Chl is the chlorophyll a concentration (µg L−1), SST is the sea surface temperature (K), SSN is sea surface nitrate (μM), and L is the latitude (°N). Hereafter, these algorithms are referred to as SD02 and W07, respectively. We compared the DMS concentrations observed during the cruise and those predicted by the SD02 and W07 parameterization schemes (Figure 2) by using the observations from the stations only (Figures 2a and 2d) and also by using the means for 1° × 1° (Figures 2b and 2e) and 0.1° × 0.1° (Figures 2c and 2f) grid cells. Our aim was to estimate the influence of sampling frequency on the reconstructions.

Figure 2.

Relationship between DMS concentrations as measured by EI-PTR-MS and values predicted by two parameterization schemes; (a–c) for SD02 and (d–f) for W07, respectively. The data are presented as discrete observations at stations in Figures 2a and 2d and as means of 1° × 1° grid squares in Figures 2b and 2e and 0.1° × 0.1° grid squares in Figures 2c and 2f. The solid lines are regression lines; the dashed lines indicate 1:1 correspondence. See text for descriptions of algorithms SD02 and W07.

[10] In comparing just the low-frequency (Figures 2b and 2e) and discrete data from stations (Figures 2a and 2d), both regression algorithms show positive correlation between the observed and predicted DMS concentrations. The W07 scheme produces the better reconstruction of the two, with the regression line being generally closer to a 1:1 correspondence (slope = 1.16 ± 0.30) and the correlation (r2 = 0.609, P < 0.05, n = 8) higher than that of SD02 (slope = 0.88 ± 0.28, r2 = 0.374, P > 0.05, n = 8). If 1°-grid means are used, the slopes of the regression lines from both algorithms (1.42 ± 0.19 for SD02 and 1.64 ± 0.16 for W07) tend to be higher than those if only station data are used. Moreover, the slopes are greater than 1 (when observed concentrations are plotted on the y axis) with statistical significance, suggesting that the use of data with high temporal resolution in the algorithms leads to underestimates. This feature is also apparent in the comparison using 0.1°-grid means (Figures 2c and 2f). Interestingly, the SD02 scheme produces estimated values that seem to be separated into two groups (Figure 2c): one just below the 1:1 line and the other far above. Most of the data in the group above the 1:1 line are from the observed DMS hot spots. In contrast, the data set from the W07 scheme falls into one group, an indication that W07 reconstructed the DMS distribution better than SD02 in this area. This is likely because the W07 scheme was developed from a DMS data set obtained in the North Pacific, whereas SD02 used global measurements. Note that W07 still underestimates the DMS concentration by ~40%, and the correlation is lower (r2 = 0.431, P < 0.05, n = 421) than that for the estimates using the 1° data set. Overall, it is clear that both algorithms tend to underestimate DMS concentrations at hot spots.

[11] There is a correlation between aqueous DMS and chlorophyll a concentrations in regions with high primary productivity such as the Southern Ocean and the North Pacific Ocean [Vallina et al., 2006] (they report r2 = 0.5–0.7 for our study area). We plotted DMS concentrations against chlorophyll a concentrations for the 0.1°-grid mean data set (Figure 3). Although there is a linear relationship between most of the DMS and chlorophyll a data, the few data from the DMS hot spots obviously deviate from the linear relationship, as they do in the comparisons with the SD02 and W07 predictions.

Figure 3.

Relationship between DMS concentrations as measured by EI-PTR-MS and chlorophyll a concentrations.

[12] These results reveal that none of the algorithms using biochemical parameters reconstructs the DMS distribution in the data set of high-frequency observations in the western North Pacific. The discrepancies are mainly for the hot spots detected by continuous measurements; the algorithms reasonably reproduce the data set from the stations, possibly because the algorithms were originally built upon discrete data sets. This result suggests that previous estimates of DMS fluxes might have been underestimated in regions of high primary productivity. For example, the DMS concentrations observed in this study were systematically higher by 40% than those predicted by W07; therefore, the flux estimates should also be 40% higher than those estimated by W07, even though there are large uncertainties caused by seasonal variations. There are probably DMS hot spots in other oceanic regions associated with high primary productivity, such as the Southern Ocean, the eastern North Pacific, and upwelling zones.

[13] We saw a clear correspondence between the variations in DMS and NCP in the study area (Figure 4). There are a high correlation for the station data (r2 = 0.944, P < 0.05, n = 7; Figure 4a) and a good correlation even for the 0.1°-grid data set (r2 = 0.676, P < 0.05, n = 279; Figure 4c). More important is the consistency of slopes and intercepts among data sets with different sampling frequencies. The slopes and intercepts are not significantly different between the two analyses inclusive of hot spots (i.e., every 1° and 0.1°). For example, the slopes for the two relationships are 0.0135 ± 0.0011 and 0.0124 ± 0.0004 for the 1° and 0.1° data, respectively (Figures 4b and 4c). These results suggest that the relationship between the distributions of DMS and NCP does not depend on the frequency of measurements. The observed clear correspondence likely indicates that the dominant factors and the time scale for the variations in DMS are similar to those for NCP. Therefore, NCP has great potential for reconstructing aqueous DMS distributions and could be a simpler parameterization than those used previously.

Figure 4.

Relationship between observed DMS concentrations and NCP. The data presented are (a) observations at discrete stations and means of (b) 1° and (c) 0.1° grids.

[14] The rate of DMS production is generally known to also depend on the phytoplankton community composition [Groene, 1995]. In contrast, NCP reflects the net biological CO2 consumption by the entire autotrophic and heterotrophic community, so that NCP variations are not dependent on the specific community composition. During the time period of our observations in the western subarctic North Pacific Ocean, haptophytes, which are known to be strong DMS producers, were predominant in the overall planktonic community, accounting for 37% ± 10% of chlorophyll a biomass at five stations north of 44°N, with heterokontophytes, green algae (chlorophytes and prasinophytes), and other phytoplankton groups each being marginal contributors of 30% ± 6%, 14% ± 4%, and 20% ± 6%, respectively (Y. Nosaka, manuscript in preparation, 2013). There were no readily apparent differences in the community composition between stations, suggesting that the dependence on community composition was negligible during the observations. This fact could help explain why DMS concentrations show a linear correlation with NCP.

[15] Our results clearly show that NCP, one of the parameters related to integrated biological activity, is an important indicator for the DMS distribution in the PSAW during summertime, and it has strong potential for application to an algorithm that reconstructs DMS distributions in the open oceans. On the other hand, at low latitudes, environmental parameters such as insolation rather than biological activities would play a key role as controlling factors [Vallina and Simó, 2007]. Therefore, the most suitable algorithm should be carefully selected for each area and season after making further observations with emerging high-frequency techniques [Saltzman et al., 2009; Kameyama et al., 2009; Asher et al., 2011], especially in areas deficient in DMS data and important for DMS flux.

Acknowledgments

[16] We express our sincere thanks to the captain and crew of R/V Hakuho Maru and all the scientists on board for their support during the SPEEDS/SOLAS 2008 cruise. This research was conducted by a Grant-in-Aid (1867001) for Scientific Research in Priority Areas “Western Pacific Air-Sea Interaction Study (W-PASS).” Funding was also provided by the Global Environment Research Fund (RFa-1102) of the Ministry of the Environment, Japan, and by a Grant-in-Aid for Scientific Research (B) (23310016) from the Ministry of Education, Culture, Sports, Science and Technology, Japan. This research is a contribution to the Surface Ocean Lower Atmosphere Study (SOLAS) Core Project of the International Geosphere-Biosphere Programme (IGBP).

[17] The Editor thanks two anonymous reviewers for their assistance in evaluating this paper.

Ancillary