Potential impact of ocean ecosystem changes due to global warming on marine organic carbon aerosols

Authors


Abstract

[1] Production of organic carbon (OC) aerosols by biological activity in the ocean is hypothesized to influence climate change. We employ model sensitivity studies to assess the effects of ocean ecosystem changes on the marine OC fluxes by using an integrated Earth system model. Our modeled estimate of global marine primary OC emission (7 Tg OC yr−1) is comparable with that calculated from sensitivity simulation which uses chlorophyll concentration constrained by the satellite observations. The model significantly underestimates OC concentration in summer over the remote ocean without the ocean source, but better agreement could be achieved by adding the marine source. We apply simulated perturbations in chlorophyll concentration and the sea-ice extent by previous model run for the IPCC AR4 to the OC source function only. Our model results suggest an insignificant change in the global mean primary OC flux from the ocean to the atmosphere but with considerably large spatial variability due to the changes in chlorophyll concentration projected by the climate warming simulation. The major features of the geographical distribution are increases over the Arctic Ocean, which are associated with the chlorophyll concentration and the sea-ice extent changes primarily driven by the enhanced light efficiency for photosynthesis in the sea-ice free ocean. In climate chemistry and aerosol models that ignore this process, the future (late 21st century) surface OC concentration would be underestimated by 40–200 ng m−3 in September over the Arctic Ocean (70–90°N).

1. Introduction

[2] It has been proposed that the marine biogeochemical sulfur cycle could stabilize the Earth's climate against perturbations by exerting a negative feedback on climate [Charlson et al., 1987]. Recently, it has been argued that marine organic carbon (OC) aerosol has a potential to modify the CLAW mechanism, because such particles could control the aerosol number concentration and change the chemical composition in the clean marine atmosphere [O'Dowd et al., 2004; Leck and Bigg, 2005; Meskhidze and Nenes, 2006; O'Dowd and de Leeuw, 2007; Andreae and Rosenfeld, 2008]. Observations of significant amounts of organic matter in submicrometer aerosols suggest that the insoluble organic aerosols are the dominant type of marine aerosols during periods of increased biological activity in the ocean [e.g., Novakov et al., 1997; Putaud et al., 2000; Cavalli et al., 2004; O'Dowd et al., 2004]. It implies a biologically driven production mechanism of potential importance to climate change. The OC is composed of water-insoluble OC (WIOC) and the water-soluble OC (WSOC). WIOC is predominantly produced via the primary OC (POC) production, while the majority of WSOC is produced via secondary OC (SOC) formation associated with photochemical oxidation processes of the volatile organic compounds (VOCs) [Ceburnis et al., 2008; Facchini et al., 2008a].

[3] During phytoplankton blooms from spring through autumn, the bursting of air bubbles produces submicron particles enriched in water-insoluble organic matter [O'Dowd et al., 2004]. Results of laboratory study provide confirmation that bursting bubbles at the ocean surface produce significant numbers of submicron organic-dominated aerosols and thus support the hypothesis that this pathway is a potentially important global source of cloud condensation nuclei [Keene et al., 2007]. However, the marine OC source is often neglected in aerosol models. Typically, models significantly underestimate OC over the remote ocean, possibly due to a missing ocean source [Tsigaridis and Kanakidou, 2003]. The top down approach suggested large sources of OC for both fine and coarse marine aerosol (8–75 Tg yr−1) from the ocean [Heald et al., 2006; Roelofs, 2008; Spracklen et al., 2008].

[4] Recently, O'Dowd et al. [2008] have developed a mechanistic model of the primary oceanic WIOC flux, which directly predicts the response of OC emissions to biological changes in the ocean, based on measurements at Mace Head (53.3°N, 9.9°W) [Yoon et al., 2007]. Subsequently, Langmann et al. [2008a] have implemented the empirical equation in a regional atmospheric climate chemistry and aerosol model. Globally, Langmann et al. [2008b] have applied their revised relationship between organic mass and chlorophyll concentration and estimated marine fluxes of submicron OC of 2.3–2.8 Tg yr−1. O'Dowd et al. [2008] have correlated average monthly chlorophyll concentration from MODIS over a fixed oceanic region (1000 km × 1000 km) upwind of observational site with the observed WIOC fraction. Spracklen et al. [2008] suggested this method could be improved by using back trajectories weighted chlorophyll concentration.

[5] Here we investigate the effects of ocean ecosystem changes on the production of OC from marine sources by using an integrated Earth System Model (ESM). We focus on the response of OC source function to changes in chlorophyll concentration and sea ice extent over the Arctic Ocean. Section 2 describes our ESM and additional processes to our ESM for this study. Section 3 describes four experiments conducted to examine how changes in the chlorophyll concentration and sea ice extent and the uncertainties in the chlorophyll concentration can affect the global oceanic emissions of OC. Section 4 examines the sensitivity of the marine OC in different simulations as well as comparisons of the model results with observations of OC. Section 5 provides a summary of our findings.

2. Model Approach

[6] For this study, the organic sea-spray source, the marine isoprene flux and the formation of secondary organic aerosol (SOA) modules have been introduced to our integrated ESM (Figure 1). We describe each component of our model in section 2.1 and the source function of marine POC in section 2.2, the isoprene flux in section 2.3, and the SOA formation in section 2.4, respectively.

Figure 1.

Schematic diagram of the integration of the marine organic carbon (OC) sources in our Earth system model. The red arrows represent the flux added in this work.

2.1. Earth System Model

[7] We use our integrated ESM, MIROC_ESM, which has been developed at the Japan Agency for Marine-Earth Science and Technology (JAMSTEC) in collaboration with the Center for Climate System Research (CCSR) of the University of Tokyo and the National Institute for Environmental Studies (NIES) [K-1 Model Developers, 2004; Kawamiya et al., 2005]. The version of MIROC_ESM used in this study is based on that described by Watanabe et al. [2008]. Our model includes atmospheric chemistry/aerosol components and marine/terrestrial biogeochemistry components on top of the atmospheric and oceanic general circulation model (AOGCM). The AOGCM is coupled without flux adjustments [K-1 Model Developers, 2004] and described for atmospheric component in section 2.1.1 and for oceanic component in section 2.1.2, respectively.

2.1.1. Atmospheric Component

[8] The atmospheric component is run with a T42 spectral resolution (2.8° × 2.8°) and 80 vertical levels. Time evolution of tracer species in the atmosphere is calculated with a time step of 20 min. The atmospheric aerosol component simulates mass concentration and optical properties of tropospheric aerosols [Takemura et al., 2000]. Simulated aerosol mass classes include OC, black carbon (BC), sea salt, sulfate aerosols and dust. The OC aerosol is currently represented by a single submicron size bin. The sea-salt flux is calculated as a function of the 10 m wind speed, based on the empirical formula of Erickson et al. [1986]. This formulation uses two sets of coefficients (slope and intercept) for moderate and high wind regimes deduced from the observations [Takemura et al., 2000, equation 5]. Since the marine-derived organic compounds could suppress surface tension, their incorporation into marine aerosols may facilitate the activation of cloud condensation nuclei into cloud droplets [e.g., Facchini et al., 1999]. However, the current version of the MIROC_ESM uses a simple relationship between the total aerosol number concentration and cloud droplet number concentration to account for both the first and second indirect effects as described by Takemura et al. [2003]. Thus more detailed model of aerosol-cloud interactions would be required to access a potential impact of marine surfactants on cloud droplet growth [e.g., Fountoukis and Nenes, 2005].

[9] We perform all the simulations under the preindustrial conditions for concentration of long-lived greenhouse gases, emissions for aerosols and O3 precursors, and the initial conditions from the equilibrium climate simulations [Nozawa et al., 2005; Watanabe et al., 2008]. Takemura et al. [2005] briefly summarized the anthropogenic emission of OC (T. Nozawa and J. Kurokawa, Historical and future emissions of sulfur dioxide and black carbon for global and regional climate change studies, manuscript in preparation, 2009) used in this study. Total anthropogenic OC emission is 7.06 Tg yr−1, which comprises from biomass burning (4.14), biofuel burning (2.23), agricultural wastes (0.36) and fossil fuel burning (0.33) [Takemura et al., 2005]. The biomass burning emission is less than a half of the estimate by Ito and Penner [2005] (9.1 Tg OC yr−1) and is likely underestimated [Ito et al., 2007a; Marlon et al., 2008]. Here, we focus on the response of marine OC to ocean ecosystem changes in the clean marine atmosphere, so that the choice of emission inventories is not expected to have a large influence on the conclusions of our sensitivity analysis.

[10] The aerosol component is coupled with an online simulation of gas-phase chemistry [Sudo et al., 2002]. Our two-way coupling between aerosols and gas-phase chemistry provides consistent chemical fields for SOA formation from gas-phase reactions of the precursors. The chemical scheme of Pöschl et al. [2000] is currently used for isoprene oxidation. The uncertainties in the chemical oxidation pathways of isoprene may lead to additional uncertainties in the impact of isoprene on SOA formation [Pöschl et al., 2000; Ito et al., 2007b, 2009; Paulot et al., 2009], and further research would be required to quantify the SOA yield from isoprene oxidation under low-NOx conditions.

2.1.2. Oceanic Component

[11] The oceanic component is simulated with a resolution of 0.5°–1.4° for latitude, 1.4° for longitude and 44 vertical levels The ocean model adopts a hybrid vertical coordinate system; the uppermost 8 levels out of the 43 use the σ coordinate and the rest the z coordinate. The vertical grid spacing varies with depth, whose values for top 100 m are 5, 5, 5, 5, 5, 5, 7.5, 7.5, 10, 10, 10, 10, 15, and 20 m from the top to the bottom. The ocean carbon cycle model is a simplified plankton dynamics model developed by Oschlies and Garçon [1999] with a series of inorganic carbon reactions recommended by Orr et al. [1999]. The ocean ecosystem model describes time evolution of nitrate, phytoplankton, zooplankton, and detritus. Phytoplankton growth depends on the local conditions of light, temperature and turbulence, and considers nitrate as the only limiting nutrient. Phytoplankton abundance is calculated on nitrogen basis in the marine biogeochemistry component of Yoshikawa et al. [2008]. Chlorophyll concentration is calculated by multiplying the nitrogen-based concentration of the modeled phytoplankton by a factor of 1.59 (gChl/molN), which corresponds to a chlorophyll to carbon mass ratio of 1:50 and a C/N mole ratio of 106/16. This conversion factor is used globally.

[12] A transient run was done for the period from 1850 to 2100 with the same protocol as used for C4MIP [Friedlingstein et al., 2006], in which the A2 Scenario of the Special Report on Emission Scenarios (SRES A2) was used for 2000–2100. Atmospheric CO2 concentration was allowed to vary as calculated by the carbon cycle components. The intensity of the positive feedback to climate change by the climate–carbon cycle model without the atmospheric chemistry component was in the middle of the range compiled by Friedlingstein et al. [2006].

2.2. Marine Primary Organic Carbon Flux

[13] We incorporate the parameterization of WIOC fraction as a function of chlorophyll concentration from O'Dowd et al. [2008] in the aerosol component of Takemura et al. [2000]. Comparison of tropical aerosol properties with those of the summer central Arctic Ocean showed similarities, suggesting that the source pattern is common over the oceans [Leck and Bigg, 2008]. Langmann et al. [2008b] revised the empirical relationship between organic mass (%) and chlorophyll concentration (mg m−3) because of their corrections of the chemical analysis. Thus the organic mass for each grid, l, and time step, t was calculated from

equation image

[14] The intercept is not given in the equation of Langmann et al. [2008b], but the value of organic mass (OM) is restricted to the range from 10% to 90% in accordance with the original formula of O'Dowd et al. [2008]. The sea-salt flux in the smallest size bin (radius range 0.1–0.316 μm) [Takemura et al., 2009, Table B1] is used for an organic source function in conjunction with the empirical relationship. Chlorophyll concentration is calculated in the marine biogeochemistry component of Yoshikawa et al. [2008].

2.3. Marine Isoprene Flux

[15] We apply marine isoprene source function to the marine biogeochemistry component of Yoshikawa et al. [2008], based on the work of Palmer and Shaw [2005]. The model assumption in the marine isoprene flux calculation is based on the observations that biological isoprene water column production and loss processes (i.e., efflux to the atmosphere, chemical degradation and biological consumption) are approximately in balance [Bonsang et al., 1992; Milne et al., 1995; Jickells et al., 2008]. The concentration of isoprene at each grid and time step in the surface ocean layer is calculated from the empirical relationship between the chlorophyll content of seawater and isoprene concentration in the ocean surface globally (R2 = 0.62) [Broadgate et al., 1997]. The ocean-to-atmosphere flux is then calculated as the product of the surface water concentration of isoprene and a piston velocity. The latter is calculated from the 10 m wind speed and the Schmidt number for the isoprene [Palmer and Shaw, 2005].

2.4. Secondary Organic Aerosol Formation

[16] The SOA formation is introduced based on the model described by Goto et al. [2008] and references therein [e.g., Chung and Seinfeld, 2002; Tsigaridis and Kanakidou, 2003]. Goto et al. [2008] have implemented a gas-particle partitioning model of SOA formation in a global three-dimensional aerosol transport radiation model [Takemura et al., 2000, 2002, 2005]. We use “monoterpene” as a proxy for all monoterpene emissions as by Goto et al. [2008] but treat the three oxidation products as a tracer. We also add a tracer of isoprene oxidation product, following Henze and Seinfeld [2006]. The oxidation of isoprene and monoterpene by the three main oxidants of OH, O3 and NO3 is taken into account in the chemistry component of Sudo et al. [2002]. The species from the oxidation path of isoprene by OH and those of monoterpene by OH, O3 and NO3 are condensed on the tropospheric OC, transported and deposited in the model.

3. Model Sensitivity Analysis

[17] We perform four experiments (Exp1–4) to investigate the effects of ocean ecosystem changes on the marine OC flux (Table 1). In order to test OC sensitivity to the variability in chlorophyll concentration and sea-ice extent, we construct four possible perturbations to OC source function only by changing the chlorophyll concentration and sea-ice extent. Here, we define a base run (Exp1) without marine OC sources as by Takemura et al. [2000]. The perturbed emissions are then implemented in our ESM, with the ocean OC emissions calculated using modeled chlorophyll concentration (Exp2), with the OC emissions calculated using observed chlorophyll (Exp3), and with the OC emissions calculated using future chlorophyll and sea-ice extent (Exp4).

Table 1. Summary of Different Experiments Performed in This Study
NumberYearChlorophyll DataMarine POC Emission ± 1σ
  • a

    Base run without marine organic carbon (OC) source.

  • b

    Sensitivity run with marine OC emission which uses the modeled chlorophyll concentration.

  • c

    The parentheses represent the annual averages only for the simulation year of 15.

  • d

    Sensitivity run with marine OC emission which uses the chlorophyll concentration constrained by the satellite observations taken from Giovanni [Acker et al., 2008].

  • e

    Sensitivity run with marine OC emission which uses the future chlorophyll concentration and sea-ice extent projected using IPCC SRES A2 emission [Yoshikawa et al., 2008].

Exp1a1–15No Effect on OC0
Exp2b1–15Model7.301 ± 0.08 (7.274)c Tg OC yr−1
Exp3d14–15Model with Observation(6.178)c Tg OC yr−1
Exp4e1–15Model with AR4 result7.356 ± 0.11 (7.329)c Tg OC yr−1

[18] Three runs (Exp1, Exp2 and Exp4) are performed for 15 years from the initial conditions in the equilibrium climate simulations [Watanabe et al., 2008; Yoshikawa et al., 2008], introducing the SOA formation mechanism from the terrestrial biosphere but without the marine source (Exp1) and with the emissions from the ocean (Exp2 and Exp4). After 5 year simulations for each experiment (Exp1, Exp2 and Exp4), the major adjustments in climate in response to the small perturbations are completed (the global mean sea-surface temperature trend for y = 6–15 ranges from −0.082 to 0.065 K yr−1). We analyze the mean state of 10 year periods (simulation years of 6–15) to estimate the perturbations due to the chaotic nature of climate model unless otherwise specified. Thus a standard deviation (σ) for the interannual variability is calculated from the monthly averaged data during ten years. All of these climate simulations do not include anthropogenic emission changes, but the effects of the emission changes on the chlorophyll concentration and the sea-ice extent are simulated in Exp4 by imposing monthly output data from the transient model simulation for the IPCC AR4 on each time step and grid.

[19] The effects of marine OC sources are examined by adding modeled OC fluxes from the ocean sources (Exp2). However, there is notable uncertainty in model estimates of chlorophyll concentration [Sarmiento et al., 2004]. Thus we conduct a sensitivity run (Exp3) with a bias correction. The data set of observed chlorophyll concentration is obtained from the Goddard Earth Sciences Data and Information Services Center (GES DISC) Interactive Online Visualization and Analysis Infrastructure (Giovanni) [Acker et al., 2008]. We use the monthly mean chlorophyll concentration from the merged data product of the MODIS (Aqua) and SeaWiFS satellite data for 2003–2007. Since the differences in the chlorophyll concentration between the model and observation vary seasonally, we have adjusted the model bias on a per month basis. We calculate the differences in the monthly averaged chlorophyll concentration between the model results of (Exp2) for the last two years and the 5 year mean chlorophyll observation. We fix the biases in the monthly averages by adding the differences to the chlorophyll concentration each time step only for source functions of POC and isoprene emissions. Thus the adjusted chlorophyll concentration, [Model]', for each grid, denoted by the subscript l, time step, t, month, m, and year, y, were calculated from

equation image

where [Model] and [Obs] are modeled and observed chlorophyll concentration, respectively. Sensitivity simulations for Exp3 are performed for two years (y = 14, 15) when the Exp2 simulations reach equilibrium states (the difference in global annual mean sea-surface temperature between Exp2 and Exp3 for y = 15, 0.27 K, is within ±1σ, 0.45 K, for y = 6–15). We analyze the last year data with the first year of simulation considered as a spin-up period. This sensitivity run would help quantify the geographical uncertainties in the marine OC emissions due to those in the chlorophyll concentration estimates.

[20] It is considered that changes in chlorophyll concentration and sea-ice extent are the major contributors to future marine OC emission changes, based on previous works focusing on the sulfur cycle [Bopp et al., 2003; Gabric et al., 2004, 2005; Gunson et al., 2006; Kloster et al., 2007]. Gabric et al. [2004] suggested the possibility that cloud microphysical changes over the Arctic Ocean resulting from the strong dimethylsulfide (DMS)-derived aerosol changes could have a significant effect on radiation budgets. Thus we apply the perturbations in chlorophyll concentration and the sea-ice extent projected by the transient model simulation for the IPCC AR4 assessments [Yoshikawa et al., 2008] to the OC source function only (Exp4). Changes in monthly averaged chlorophyll concentration and sea-ice extent at each grid were obtained from the climate warming experiments using IPCC SRES A2 emission by Yoshikawa et al. [2008]. The averaged chlorophyll concentration in September increased by a factor of 9 from the preindustrial time to the future over the Arctic region, as the sea ice extent decreased to zero (Figure 2). Definition of the Arctic region used in our analysis is the areas covered by sea ice in 1850. We note that the changes in other regions, where the ocean is not covered by the ice in 1850, are much smaller than those in the Arctic region. We choose the period 1850–1870 for the preindustrial conditions and period 2080–2100 for the future conditions. We use the monthly averaged chlorophyll concentration difference between these two periods simulated by Yoshikawa et al. [2008] to estimate a future chlorophyll concentration change. We have added this to the chlorophyll concentration at each grid and time step, but the sum is used only for source functions of POC and isoprene emissions. Thus the chlorophyll concentration change does not affect other physical and chemical variables directly but can influence only through the small feedback from the climate change induced by changes in the OC concentration on oceanic biological production [Gunson et al., 2006]. Similarly, the sea-ice extent changes are only taken into account for the POC source function, but are not applied to the isoprene emissions. The organic sea-spray flux may also be affected by the wind changes, but the effect of the changes would be minor. For instance, Stier et al. [2006] found no significant trend in the sea salt emission and only minor variations in the total aerosol burden and meridional distribution from 1860 to 2100.

Figure 2.

Simulated changes in modeled chlorophyll concentration (mg m−3) (red) and ice extent (106 km2) (black) in September from the preindustrial time to the future over the Arctic region by Yoshikawa et al. [2008]. The Arctic region used in this analysis is the sea-ice covered areas in 1850.

4. Model Results and Discussion

4.1. Global Estimates of Marine Organic Carbon and Isoprene Emissions

[21] The estimates of global submicron OC emissions from the ocean are highly uncertain, ranging from 2.3 to 5.5 Tg OC yr−1 [Langmann et al., 2008b; Spracklen et al., 2008; Gantt et al., 2009]. The top down estimate (5.5 Tg OC yr−1) by Spracklen et al. [2008] is comparable to our modeled estimate of global marine POC emission (Exp2, 7.3 Tg OC yr−1) as well as that of Exp3 (6.2 Tg OC yr−1) which uses chlorophyll concentration constrained by the satellite observations. Here, we compare the two experiments for the same simulation year (y = 15 in equation (2)).

[22] Globally, both results of the estimated POC emissions using modeled chlorophyll concentration (Figure 3a) and observed chlorophyll concentration (Figure 3b) show similar geographical patterns, which primarily reflect the sea-salt fluxes and the ocean biomass. Results of Exp2 show significant OC emissions (0.5–2.5 ng C m−2 s−1) in the North Atlantic, North Pacific and the Southern Ocean, which are consistent with those of (Exp3). These geographical features are associated with the distribution of the high-biomass regions. However, our modeled OC emissions of (Exp2) are generally larger than those of (Exp3). A possible reason for our overestimates in chlorophyll concentration would be the deficiency of multielement nutrient limitation such as iron, phosphate, and silicate, which is important over significant portions of the surface ocean, particularly in the high-nitrate, low-chlorophyll (HNLC) regions [e.g., Martin, 1990; Moore et al., 2002]. Further work is required to improve both the ocean biogeochemical model and aerosol model and to refine the organic-inorganic sea-spray flux formulation.

Figure 3.

Global distribution of annual mean oceanic emission of primary OC (POC) (ng C m−2 s−1) estimated using (a) modeled chlorophyll concentration (Exp2), and (b) observed chlorophyll concentration (Exp3).

[23] The estimates of global marine isoprene emissions are also uncertain, ranging from 0.1 to 1.9 Tg yr−1 [Bonsang et al., 1992; Broadgate et al., 1997; Palmer and Shaw, 2005; Arnold et al., 2009; Gantt et al., 2009]. Our estimate of a marine isoprene source of 0.1 TgC yr−1 is at the lower end of the range, but is in good agreement with the estimate using remotely sensed chlorophyll observation by Palmer and Shaw [2005]. Arnold et al. [2009] estimated 0.04 TgC yr−1 of SOA from the top down isoprene source (1.9 Tg yr−1) by applying the 2% SOA yield from isoprene globally [Henze and Seinfeld, 2006]. This is equal to 0.6% of the marine POC (7.2 Tg yr−1) in our model. Thus the contribution of SOC production from oceanic isoprene to total oceanic OC emission would be small even if the higher end of isoprene emission is applied to our model.

4.2. Global Distribution of Organic Carbon Concentration

[24] We show the differences in estimated OC concentration of (Exp2)–(Exp1) between with and without the emissions from the ocean in Figure 4. In January (Figure 4a), our oceanic OC emissions enhance surface OC concentration (100–200 ng m−3) due to the inclusion of the OC source in the Southern Ocean (30–60°S). In July (Figure 4b), the modeled OC is enhanced in the Northern Hemisphere with the largest enhancement (100–300 ng m−3) in the North Atlantic and North Pacific Ocean. The seasonal variation of modeled OC concentration at high latitudes with a minimum in winter and a maximum in summer is generally consistent with that of measurement owing to the OC source driven by oceanic biological material. We note that Figure 4 shows continental hot spots (e.g., Africa, Southeast Asia, and South America) in OC concentration calculated from the difference with and without marine OC sources, but they are much smaller than marine OC on a percentage basis.

Figure 4.

Geographical distributions of the differences in OC surface concentration (ng m−3) with and without the emissions from the ocean for (Exp2)–(Exp1) in (a) January and (b) July.

[25] Observational data for OC at marine sites are very limited. Comparisons of monthly averages between model results and measurements are quite problematic at marine sites near the continents, because the coarse horizontal model resolution does not resolve the distinctive local pattern of phytoplankton blooms along the shoreline. Here, model performance is evaluated by comparing the estimated OC concentration with observations at seven sites as background conditions over the open ocean: Putaud et al. [2000] (28.3°N, 16.5°W), Phinney et al. [2006] (50.0°N, 145.0°W), Virkkula et al. [2006] (38.56–57.52°S, 5–15°E), Pio et al. [2007] (38.7°N, 27.4°W), Yoon et al. [2007] (53.3°N, 9.9°W), Zorn et al. [2008] (40–45°S, 55–60°W), and Sciare et al. [2009] (37.5°S, 77.3°E) (Table 2 and Figure 5). The model significantly underestimates OC over the open ocean without the ocean sources (Exp1), but better agreement can be achieved by adding the marine POC source. Figure 4 shows a remarkable enhancement in OC concentration during January over the South Pacific Ocean. We were not able to find literature data on OC concentration in this region. The OC concentration (120–200 ng m−3) may be comparable to the measurement of 243 ng m−3 by Zorn et al. [2008], but, of course this is not a direct comparison at the same location, only tells the magnitude is not unrealistic. These results suggest that measurements in these regions may provide constraints on the marine OC sources.

Figure 5.

Seasonal cycle of observed total OC aerosol ± 1σ (except for months with a single value) (green lines), observed WIOC ± 1σ (except for months with a single value) (black lines), and modeled OC ± 1σ (Exp2, red lines) at (a) Mace Head (53.3°N, 9.9°W) [Yoon et al., 2007] and (b) Amsterdam Island (37.5°S, 77.3°E) [Sciare et al., 2009]. The total OC at Amsterdam Island represents the non-methanesulfonate OC which is not originated from the oxidation of dimethylsulfide.

Table 2. Comparison of Simulated and Observed OC Concentration ± 1 σ over the Ocean
StudyMonthLocationObservation (ng m−3)Exp1 (ng m−3)Exp2 (ng m−3)
  • a

    December, January, and February.

  • b

    Water-insoluble OC (WIOC) is used for the comparison. The parentheses represents the total OC. The WIOC mass concentration was calculated as the difference between OC and WSOC mass concentration.

  • c

    June, July, and August.

  • d

    A conversion factor of 1.4 for OM/OC is used.

Putaud et al. [2000]July28.3°N, 16.5°W210 ± 120103 ± 16216 ± 20
Phinney et al. [2006]July50.0°N, 145.0°W300 ± 1005 ± 199 ± 11
Virkkula et al. [2006]January38.56–57.52°S, 5–15°E88 ± 463 ± 183 ± 25
Pio et al. [2007]DJFa38.7°N, 27.4°W89 (270)b7 ± 552 ± 21
Pio et al. [2007]JJAc38.7°N, 27.4°W169 (380)b25 ± 8113 ± 24
Zorn et al. [2008]January40–45°S, 55–60°W243d24 ± 6161 ± 50

[26] We compare total OC aerosol and WIOC with the observed seasonal cycle at two marine locations (Figure 5). Observations are taken from Mace Head [Yoon et al., 2007] and Amsterdam Island [Sciare et al., 2009]. The OC concentration in model with the ocean source (Exp2) is in good agreement with observation of WIOC at Mace Head (annual mean difference is −26%), while the standard deviation in the monthly average is significantly large, reflecting that high OC concentration is sensitive to the event of high organic matter concentrated at the ocean surface, which could result from plankton blooms. The OC in model with the ocean source (Exp2) is in fairly good agreement with observations of WIOC at Amsterdam Island (annual mean difference is −16%), while the total OC is underestimated even in winter. Spracklen et al. [2008] obtained better match with the observation by adjusting the total (primary and secondary) oceanic OC source. Our modeled OC over the remote ocean is mainly produced via the POC production, because our SOC from isoprene oxidation products is negligible as described in section 4. These results may imply additional formation of SOA which is not taken into account in the model. Recently, Facchini et al. [2008b] suggested that amines (dimethyl- and diethylammonium salts) represent a significant source of SOA (9% ± 5%) in the period of high biogenic activity. Nonetheless, a quantitative understanding of the observed high concentration of oxidized organic matter in marine aerosol remains elusive.

4.3. Potential Organic Carbon Increases in Future Chlorophyll Conditions

[27] A number of global modeling studies have suggested that the shrinking of the Arctic sea ice cover could lead to near sea ice free Septembers by around the mid-21st century [Holland et al., 2006]. Retreat of Arctic summer sea ice would result in more sunlight for photosynthesis and greater surface warming, which may lead to enhanced oceanic biological production [Gabric et al., 2005]. Thus we compare the marine primary OC emissions of Exp2 and Exp4 in September. Globally, our model results suggest an insignificant change (within ±1σ) in the annual mean marine OC flux to the atmosphere due to the changes in chlorophyll concentration projected by the climate warming simulation. Jacobson and Streets [2009] included a fixed value of the content of primary organic matter in the sea spray function, which depended on wind speed, temperature, and sea ice. They found similar changes of the sea spray (+0.68%) and ocean bacteria (+0.72%) emissions due to the effect of future anthropogenic emissions (SRES B1) on climate for 2030, and the resulting feedback to natural emissions. However, they did not consider the effect of biogeochemical changes on OC. In climate chemistry and aerosol models that ignore this process, the regional surface OC concentration would differ significantly.

[28] The geographical distributions of the differences in marine primary OC emissions between Exp2 and Exp4 show large regional variability (Figure 6). Our results show small decreases (up to −0.5 ng C m−2 s−1) in the OC emissions over most of midlatitude and low-latitude oceans (60°N–60°S). Enhanced stratification of the upper ocean and decreased intensity of tropical upwelling reduced the oceanic primary production through reduced nutrient supply. This leads to the decrease in the chlorophyll concentration and thus the decrease in the OC emissions over of midlatitude and low-latitude surface oceans. Here, we find significant increases (up to 1.5 ng C m−2 s−1) at higher-latitude regions. Enhanced stability and reduced vertical exchanges increased the oceanic primary production in higher-latitude regions through increased light efficiency for photosynthesis due to a longer growing season. This leads to the increase in the chlorophyll concentration and thus the increase in the OC emissions over the polar and subpolar ocean.

Figure 6.

Geographical distribution of the differences in oceanic emission of POC (ng C m−2 s−1) for (Exp4)–(Exp2) in September.

[29] The potential contribution to OC concentration from the ocean source of (Exp4)–(Exp1) in the future September is estimated from the differences with and without the marine POC source (black line in Figure 7). The marine OC is highly sensitive to chlorophyll concentration and the sea-ice extent in the Arctic Ocean. Our results suggest that the future surface OC concentration is underestimated by 40–200 ng m−3 (zonal average: 47–87 ng m−3) in September over the Arctic Ocean (70–90°N) in climate chemistry and aerosol models which ignore the marine POC source. It is noteworthy that the difference with and without the marine source over the Arctic Ocean (black line in Figure 7) is practically identical to the increase from preindustrial conditions for chlorophyll concentration (Exp2) to the future (Exp4) (red line in Figure 7), because the ocean is covered by the ice under preindustrial conditions, which means zero OC emissions from the ocean (see Figure 3).

Figure 7.

Zonal mean of the differences in OC concentration (ng m−3) for (Exp4)–(Exp1) (black), and for (Exp4)–(Exp2) (red) over the ocean for September.

5. Conclusions

[30] We investigated the effects of ocean ecosystem changes on marine organic carbon aerosols by using our integrated ESM. We introduced the mechanistic model that represents marine sources of organic aerosols to our ESM, based on the work of O'Dowd et al. [2008] for POC, Palmer and Shaw [2005] for marine isoprene flux, and Henze and Seinfeld [2006] for SOA formation from the isoprene, respectively. Our modeled estimate of global marine primary OC emission (7 Tg OC yr−1) was in good agreement with that calculated from sensitivity simulation which used chlorophyll concentration constrained by the satellite observation. Although the former was larger than the latter in some regions, our model generally captured geographical features, which were associated with the distribution of the high biomass and strong wind regions. The modeled overestimate of chlorophyll concentration might be associated with the influence of aerosol supply of nutrients such as iron, phosphate, and nitrate, which would be important over much of the surface ocean, particularly in the high-nitrate, low-chlorophyll regions [e.g., Fung et al., 2000; Mahowald et al., 2005; Luo et al., 2008; Solmon et al., 2009]. Gunson et al. [2006] suggested that the negative feedback of climate change on ocean DMS production could be enhanced by the inclusion of dust deposition, while Kloster et al. [2007] concluded that the overall impact of the changes in the dust deposition on the DMS production in their simulation was only of minor importance. More work is necessary to fully understand the source change processes due to the interactions between the ocean ecosystem and atmospheric aerosols.

[31] The model without the ocean source significantly underestimated OC in summer over the remote ocean, but better agreement could be achieved by adding the marine source. Our OC was in good agreement with the observation of WIOC, while our OC was lower than the measurement of total OC at several locations. The biogenic organic aerosols from sea-spray source should be included in the ESM, but further work is required to develop numerical models for the secondary organic aerosols from ocean for an accurate representation of background organic marine aerosols.

[32] Our model results suggest considerably large spatial variability in the monthly mean primary OC emissions due to the changes in chlorophyll concentration projected by the climate warming simulation. The major features of the geographical distribution in the OC concentration increases are found over the Arctic Ocean. The climatic consequences of the increased OC in the Arctic could be significant. The fine sea spray particles, which may be dominated by organics, would be important for controlling aerosol number concentration and offer surface and mass on which SOA precursor gases and DMS may condense over the Arctic Ocean. Our results imply that future OC increases in the Arctic open oceans may contribute to a negative aerosol-cloud feedback, which in turn could possibly induce an ice/snow albedo-temperature feedback. However, the strength of the feedback is a complicated function of the initial extent of the sea ice and the interactions between the atmosphere and ocean in the Polar Regions. Further work is required to investigate the impact of the OC in the atmosphere on cloud properties and sea-ice covers.

Acknowledgments

[33] We thank the contributors to the development of the MIROC_ESM. Support for this research was provided by Innovative Program of Climate Change Projection for the 21st Century (MEXT). All of the global simulations were performed using the Earth Simulator.