Based on simulations using the University of Victorias Earth System Climate Model, we analyzed the responses of the ocean carbon cycle to increasing atmospheric CO2 levels and climate change from 1800 to 2500 following the RCP 8.5 scenario and its extension. Compared to simulations without climate change, the simulation with a climate sensitivity of 3.0 K shows that in 2100, due to increased atmospheric CO2 concentrations, the simulated sea surface temperature increases by 2.7 K, the intensity of the North Atlantic deep water formation reduces by 4.5 Sv, and the oceanic uptake of anthropogenic CO2 decreases by 0.8 Pg C. Climate change is also found to have a large effect on the North Atlantics ocean column inventory of anthropogenic CO2 . Between the years 1800 and 2500, compared with the simulation with no climate change, the simulation with climate change causes a reduction in the total anthropogenic CO2 column inventory over the entire ocean and in North Atlantic by 23.1% and 32.0%, respectively. A set of simulations with climate sensitivity variations from 0.5 K to 4.5 K show that with greater climate sensitivity climate change would have a greater effect in reducing the oceans ability to absorb CO2 from the atmosphere.
Climate change ; Ocean carbon cycle ; Carbon cycle modeling
According to the Intergovernmental Panel on Climate Change (IPCC) Fifth Assessment Report (IPCC, 2013 ), many currently observed climate changes are unprecedented and the existence of global warming since the 1950s is unequivocal. Due to human activities such as fossil fuel use and land use changes since pre-industrial times, CO2 concentrations in the atmosphere have increased by about 40%. Also, only about half of the total anthropogenic CO2 emissions remain in the atmosphere ((240 ± 10) Pg C); the rest is absorbed by land ((160 ± 90) Pg C) and the ocean ((155 ± 90) Pg C) (IPCC, 2013 ). Based on simulations using 11 coupled climate–carbon cycle models, Friedlingstein et al. (2006) concluded that climate change will reduce the efficiency of the land and the ocean in absorbing atmospheric CO2 . Furthermore, the IPCC reported that CO2 increases and climate change have different effects on the lands and oceans carbon storage. CO2 increases will lead to the increased carbon storage by the land and ocean. Climate change will reduce the land and oceans capacity to absorb atmospheric CO2 due to the increasing temperatures of both the land and ocean and increasing oceanic stratification. The ocean, which has absorbed 27.9% of the anthropogenic CO2 in the past 200 years (IPCC, 2013 ), plays a crucial role in the global carbon cycle.
The ocean carbon cycle is determined by a series of complex interactions involving the air-sea exchange, inorganic carbon chemistry, ocean general circulation, and marine biological processes (Plattner et al., 2001 ). Recently, many studies have discussed the ocean carbon cycle impact of increased atmospheric CO2 and climate change. These studies indicate that rising atmospheric CO2 levels lead to increased radiative forcing produced by greenhouse gases (IPCC, 2013 ), which causes sea surface warming. Warming seawater will decrease the CO2 solubility of seawater (Plattner et al., 2001 ), resulting in a reduction in oceanic CO2 uptake. Furthermore, global warming will cause increased ocean stratification (Sarmiento et al., 1998 ), which then will cause a reorganization of the thermohaline circulation and the reduction or even collapse of the North Atlantic deep water formation (Manabe and Stouffer, 1994 and Stocker and Schmittner, 1997 ). These changes will reduce the total transport of oceanic CO2 from the ocean surface to the deep ocean (Maier-Reimer et al., 1996 and Matebr and Hirst, 1999 ). The reduction in oceanic carbon uptake, caused by the processes mentioned above, will in turn accelerate the increase in atmospheric CO2 (Joos et al., 1999 ). Ocean carbon cycle studies in China began later than other studies worldwide, but have developed rapidly. The early studies mainly used two-dimensional carbon cycle models. For example, Xu et al. (1997) discussed the inorganic carbon cycle, and the sinks and sources of oceanic carbon uptake. Pu and Wang, 2000 and Pu and Wang, 2001 analyzed the distribution of chemical components related to carbon and the critical factors influencing the distribution of carbon in the Indian Ocean. At present, three-dimensional ocean carbon cycle models are being widely used in climate change research. Xing (2000) and Jin and Shi (2001) pointed out that the biological pump plays an important role in the ocean carbon cycle. Li and Xu (2012) used a perturbation approach to compare the different effects on oceanic CO2 uptake in the Pacific Ocean and also modeled its biological processes. In addition to discussing the factors that influence oceanic CO2 uptake, other studies (Xu and Li, 2009 and Bao et al., 2012 ) have analyzed oceanic CO2 uptake and distribution based on simulations. Moreover, Wei et al., 2012 and Wei et al., 2014 employed an earth system model (BNU-ESM) to investigate the different climate change and ocean warming responses, including oceanic CO2 uptake, carbon sequestration, and ocean acidification, to various anthropogenic CO2 emission scenarios.
Based on the previous work noted above, this study uses an earth climate model to simulate the effect of increased atmospheric CO2 and associated climate changes on oceanic CO2 uptake and the ocean carbon cycle.
The University of Victoria Earth System Climate Model (UVic) was used, which consists of an energy-moisture balance atmospheric model (Fanning and Weaver, 1996 ), a dynamic-thermodynamic sea-ice model (Bitz et al., 2001 , Hibler, 1979 and Hunke and Dukowicz, 1997 ), and a primitive equation ocean general circulation model (Pacanowski, 1995 ). This model has a horizontal resolution of 3.6° longitude and 1.8° latitude, and divides the ocean into 19 vertical layers.
In order to guarantee the models computational efficiency, we used a simplified atmospheric model with only one layer, and included only CO2 forcing, with no other greenhouse gas forcing such as CH4 or aerosol forcing. The vertically integrated thermodynamic energy balance equations assume that the energy and specific humidity decrease vertically with specified scale heights. Momentum conservation equations are replaced by specified wind data, and the atmospheric heat and moisture transport by diffusion are parameterized for simplification (Weaver et al., 2001 ). The models wind stress data, 1958–1998 daily reanalysis data (Kalnay et al., 1996 ), are used to force the ocean and ice components and to calculate the latent heat and sensible heat fluxes between the atmosphere and ocean or ice components. The coupled models ocean component is Modular Ocean Model (MOM) version 2.2, which is based on Navier Stokes equations that are conditional on Boussinesq and hydrostatic approximations. In the sea-ice component, an elastic-viscous-plastic rheology represents the sea-ice dynamics, and various options for the sea-ice thermodynamics and thickness distribution are included. More information and equations may be found in Weaver et al. (2001) .
In addition, our model simulations of the ocean carbon cycle are based not only on the inorganic carbon cycle, following the protocol of the Ocean Carbon Cycle Model Intercomparison Project (Orr et al., 1999 ), but also take into consideration the ocean ecosystem, including the interactions between nutrients, phytoplankton, zooplankton, and detritus (Schmittner et al., 2008 ). Model simulations of the terrestrial carbon cycle and vegetation are based on the Met Office-Hadley Centre TRIFFID dynamic vegetation model (Meissner et al., 2003 ). The model we used in this paper simulates well the large-scale distribution of key climate variables (Weaver et al., 2001 ) and the ocean carbon cycle (Schmittner et al., 2008 ), and has been widely used in various global climate and carbon cycle studies (Friedlingstein et al., 2006 and Schmittner et al., 2008 ).
Before conducting the sensitivity simulations, we first set the atmospheric CO2 concentration at the pre-industrial level of 280 × 10−6 , and set the mean state of the climate from 1960 to 1990 as the initial condition (Weaver et al., 2001 ). Then we spun up the model for 10,000 years to reach a quasi-equilibrium state between the climate and carbon cycle. At this time, the air-sea carbon flux is close to zero. Then we conducted transient simulations based on specified atmospheric CO2 levels. Starting from the end of the spin-up simulation, the atmospheric CO2 concentration is based on observations made from 1800 to 2005, and under the RCP 8.5 scenario after 2005. According to the IPCC (2013) , the Representative Concentration Pathways (RCPs) usually refer to the pathways extending to 2100 and their corresponding emission scenarios. Extended Concentration Pathways (ECPs) describe the RCPs' extensions from 2100 to 2500. Specifically, RCP 8.5 is a relatively high CO2 concentration pathway; the atmospheric CO2 concentration reaches to 935.9 × 10−6 by year 2100, then continues to rise over time, and the corresponding ECP maintains a constant concentration after 2250 (Fig. 1 a).
Time series with (solid lines) and without (dashed lines) climate change, (a) atmospheric CO2 , (b) global mean surface air temperature change (relative to 1800), (c) global mean sea surface temperature change (relative to 1800), (d) intensity of the North Atlantic deep water formation (NADW), (e) global mean surface DIC concentration, (f) global mean surface alkalinity, (g) global mean surface salinity, (h) oceanic uptake of anthropogenic CO2 .
Based on the specified atmospheric CO2 concentration, two simulations were performed. One simulation considers climate change due to atmospheric CO2 increase and the impact on the ocean carbon cycle due to climate change. The other simulation does not consider climate change due to the atmospheric CO2 increase, but assumes that while the change in atmospheric CO2 has no direct radiation effect on global climate, the atmospheric CO2 increase has a small effect on global warming due to its influence on terrestrial vegetation. In the first simulation, we specified different climate sensitivity values. The equilibrium climate sensitivity refers to the change in the annual and global mean surface temperature when the climate system reaches a steady state, following a doubling of the atmospheric CO2 concentration. Empirical equations are adopted in the model to calculate the radiative forcing by the change in the atmospheric CO2 concentration, and therefore radiative forcing can be modified by changing the empirical equations' coefficient, thus changing the climate sensitivity. In our simulations, the climate sensitivity ΔT2x varies from 0.5 K to 4.5 K at intervals of 0.5 K (ΔT2x = 0.5 K, 1.0 K, 1.5 K, 2.0 K, 2.5 K, 3.0 K, 3.5 K, 4.0 K, and 4.5 K). Current estimates of climate sensitivity are 1.5–4.5 K, according to IPCC Fifth Assessment Report (IPCC, 2013 ). In this study, we focus on the difference between the simulation with a climate sensitivity of 3.0 K and the simulation with no climate change with respect to the ocean carbon cycle and the oceans ability to uptake atmospheric CO2 . We also analyze the possible responses of the ocean carbon cycle to different climate sensitivities.
Given a specified CO2 concentration (RCP 8.5, Fig. 1 a), the responses of the oceanic CO2 uptake to increasing atmospheric CO2 concentrations and/or climate change are as follows. If increased atmospheric CO2 has no influence on climate change through radiation, the changes in the global mean land surface air and sea surface temperatures are small enough to be ignored (Fig. 1 b and c), and can be regarded as resulting in no climate change (climate sensitivity is 0). When the climate sensitivity is set to 3.0 K, global mean land surface air and sea surface temperatures show a rising trend. Even if the CO2 concentration remains constant after the year 2250, these temperatures still continue to rise at a relatively slower rate (Fig. 1 b and c). Relative to the year 1800, the global mean land surface air and sea surface temperatures increase by nearly 8.6 K and 6.6 K by the year 2500, respectively. Research has indicated that, under the same salinity, alkalinity, and dissolved inorganic carbon concentrations, a sea surface warming of 1 K will increase the sea surface CO2 partial pressure about 4% (Takahashi et al., 1993 ). Meanwhile, sea surface warming will reduce seawater CO2 solubility (Plattner et al., 2001 ), thus decreasing the oceanic CO2 uptake. As shown in Fig. 1 e–g, compared with the simulations with no climate change, decreases in the global mean surface dissolved inorganic carbon (DIC) concentrations, alkalinity, and salinity by year 2500 are 83.4 μmol kg−1 (3.7%), 57.2 μmol kg−1 (2.5%), and 0.2 psu (0.4%) respectively, all of which reach maximum values. In addition, the changes in DIC concentrations and alkalinity have opposite effects on sea surface CO2 partial pressure whereby one can nearly offset the other (Plattner et al., 2001 ). Moreover, Zhou et al. (2005) reported that global warming leads to a rise in sea surface temperature and a dilution in sea surface water at the high latitudes of the North Atlantic Ocean. This will reduce the density of seawater and weaken the meridional density gradient between the high and low latitudes, which will result in the weakening of the thermohaline circulation. This weakening will mainly be limited to the North Atlantic rather than the whole conveyor belt. As shown in Fig. 1 d, in our simulation, compared with the no climate change case, the intensity of the North Atlantic deep water formation, represented by the maximum meridional overturning stream function, decreases by about 7 Sv at most (1 Sv = 106 m3 s−1 , year 2185). The weakening of the North Atlantic deep water formation affects the transport of carbon from the sea surface to deep water, thereby influencing the oceanic CO2 uptake (Friedlingstein et al., 2006 ). Since the rate of oceanic CO2 uptake depends on CO2 solubility, the differences between atmospheric and sea surface CO2 partial pressure, and the ocean dynamic transport of DIC, climate change reduces the oceanic uptake of anthropogenic CO2 emissions, and the specific tendencies over time are shown in Fig. 1 h. In our simulations, the oceanic uptake of anthropogenic CO2 (this years total oceanic CO2 uptake minus that in 1800) reaches a maximum of 6.0 Pg C in 2093 with no climate change, while it reaches a maximum of 5.3 Pg C in 2083 with climate change. And the maximum difference of 1.2 Pg C between the two simulations occurs in 2180.
In order to investigate the distribution of anthropogenic CO2 in the ocean, we vertically integrated the DIC concentration change (DIC concentration in a given year minus that in 1800) in all 19 layers to produce a global distribution map of the anthropogenic CO2 column inventory with and without climate change, as shown in Fig. 2 . Anthropogenic CO2 is not uniformly distributed throughout the ocean. Whether or not there is climate change, the maximum vertically integrated anthropogenic CO2 concentration is always found in the North Atlantic Ocean. Due to the formation of North Atlantic deep water, this basin is a major area where seawater sinks (Stocker et al., 1994 and Sabine et al., 2004 ). As it sinks, this seawater can transport the anthropogenic CO2 absorbed by the ocean surface to depth more efficiently than in other basins, and thus accumulates a large concentration of anthropogenic CO2 . According to our model, the North Atlantic (north to the equator) and the Arctic Ocean account for 10% of the global ocean area and 11.9% of the global ocean volume, but they stored 24.0% (without climate change) and 23.5% (with climate change) of the global oceanic uptake of anthropogenic CO2 in 2010, respectively, and 22.0% and 19.5% in 2500, respectively. In 2010, there were no obvious differences in the vertically integrated anthropogenic CO2 concentrations between the simulations with and without climate change; the difference in the column inventory of anthropogenic CO2 is just 4.1 Pg C (Table 1 ). However, as atmospheric CO2 is absorbed continuously by the ocean and is then transported to the deep ocean, the difference becomes more significant (Fig. 2 b and c). By 2500, compared to the simulation with no climate change, climate change results in a decrease of 23.1% in the column inventory of anthropogenic CO2 in the entire ocean, and of 32.0% in the North Atlantic-Arctic ocean (Table 1 ).
Column inventory of anthropogenic CO2 (the difference: with climate change minus without climate change, unit: mol m−2 ; the percentage: the difference in the results between the simulation with and without climate change divided by the result from the simulation without climate change, unit: %).
|Year||Entire ocean||Atlantic-Arctic ocean|
|No climate change||ΔT2x = 3.0 K||Percentage||No climate change||ΔT2x = 3.0 K||Percentage|
Next, we divided the ocean into the Atlantic-Arctic and the Pacific-Indian basins and analyzed the latitude-depth distribution of anthropogenic CO2 to gain a better understanding of the distribution and transport of anthropogenic CO2 .
Anthropogenic CO2 enters the ocean via the air-sea exchange and the contours of the highest anthropogenic CO2 concentrations appear in basins shallower than 500 m under each scenario (Fig. 3 ). This means that anthropogenic CO2 concentrates in the near surface water, and the thermocline is a crucial area for the oceanic uptake of anthropogenic CO2 . As shown in Fig. 3 , the highest concentration of anthropogenic CO2 is found in the subtropics where near surface seawater shows convergence and subsidence, thus allowing DIC to be transported to the deep ocean, and benefitting further oceanic uptake of anthropogenic CO2 (Bao et al., 2012 ).
Latitude-depth distributions of anthropogenic CO2 (in μmol kg−1 ) in years 2010, 2100, and 2500 for the Atlantic-Arctic (a, b) and Pacific-Indian oceans (c, d) with climate change (a, c) and without climate change (b, d).
The penetration depth of anthropogenic CO2 in the ocean depends on the rate that the near sea surface anthropogenic CO2 is transported to the interior (Sabine et al., 2004 ). And the transport rate mainly depends on ocean circulation factors, such as thermocline ventilation, deep water and intermediate water formation, and so on (Friedlingstein et al., 2006 ). As it is affected by these factors, the penetration of anthropogenic CO2 absorbed by the Pacific-Indian ocean is shallower and slower, and the penetration of anthropogenic CO2 absorbed by the basins with deep water formation, such as the North Atlantic Ocean, is deeper and faster (Fig. 3 ). One study reports that regardless of whether basins are characterized by deep water formation or not, the time scales of the near surface water being mixed to the deep ocean may be centuries (Sabine et al., 2004 ). Climate change increases sea surface temperature (Fig. 1 b), which causes a decrease in the oceanic CO2 uptake by reducing the seawaters CO2 solubility (Plattner et al., 2001 ). This also leads to increased stratification (Sarmiento et al., 1998 ). The increased stratification leads to a decrease in vertical mixing along the isopycnals and overturning circulation, which decreases the carbon vertical transport. The increased stratification also causes a gradual collapse of thermohaline circulation throughout the entire deep ocean — the thermohaline circulation plays the major role in the carbon balance on a century time scale (Sarmiento et al., 1998 ), and its collapse has been verified in related early simulations (Joos et al., 1999 ). As Fig. 3 shows, in the year 2500 the 240 μmol kg−1 contour of the DIC concentration can extend to a depth of 4,000 m in the Atlantic-Arctic ocean without climate change (Fig. 3 a), but in the simulation with ΔT2x = 3.0 K, it is confined to the sea surface in most of the world ocean (Fig. 3 b).
To further analyze the effects of climate change on the ocean carbon cycle, we calculated changes in the global oceanic anthropogenic CO2 uptake for the years 2100, 2500, and 1800–2100, as well as the 1800–2500 cumulative uptake for different climate sensitivities. The climate sensitivities range from 0.5 K to 4.5 K at intervals of 0.5 K, and with no climate change the climate sensitivity is 0, as shown in Fig. 4 . In 2100, the oceanic uptake of anthropogenic CO2 shows a linear change along with the climate sensitivity. When climate sensitivity increases by 0.5 K, the annual uptake decreases by about 0.1 Pg C, and cumulative uptake decreases by about 6.5 Pg C. In 2500, the change does not display a linear trend. Without climate change, the oceanic anthropogenic CO2 uptake is 1.5 Pg C and the cumulative uptake is 1,996.0 Pg C. When the climate sensitivity reaches 4.5 K, the annual uptake is 0.9 Pg C and the cumulative uptake is 1,502.8 Pg C, representing decreases of 39.9% and 24.7%, respectively. As shown in Table 2 , both the global mean surface air warming and global mean ocean warming values become larger as climate sensitivity increases. Regarding the global mean ocean warming, simulated under different climate sensitivities (horizontal axis), the oceanic carbon uptake and cumulative uptake of anthropogenic CO2 show a similar trend in the Figures whose horizontal axis is climate sensitivity. This series of simulations illustrate that climate warming decreases oceanic carbon uptake, and the higher the climate sensitivity or the greater the ocean warming, the larger is the reduction in the oceanic uptake of anthropogenic CO2 .
Oceanic uptake of anthropogenic CO2 for years 2100 and 2500, and the cumulative anthropogenic CO2 uptake (1800–2100 and 1800–2500) as a function of climate sensitivity (a–d) and global mean ocean warming (e–h) (Each dot represents a specified simulation with a specified climate sensitivity ΔT2x = 0,0.5, 1.0, 1.5, 2.0, 2.5, 3.0, 3.5, 4.0, 4.5 K).
|Surface air||Entire ocean||Surface air||Entire ocean|
In this study, we chose a specified atmospheric CO2 concentration to simulate resultant changes in the oceanic CO2 uptake and the ocean carbon cycle affected by climate change. The ocean is an important carbon sink. In reality, if the CO2 uptake by the ocean decreases, the anthropogenic CO2 accumulating in the atmosphere will increase, which will in turn affect climate change. Therefore, it is necessary to study the impact of climate change on the ocean carbon cycle and the resulting feedback of the ocean carbon cycle to climate change. In future studies, we will use a specified CO2 emission to investigate the feedback interactions between climate change, atmospheric CO2 concentrations, and the ocean carbon cycle.