Effects of solar radiation modification on the ocean carbon cycle: An earth system modeling study

Xiaoyu Jin, Long Cao , Jingyu Zhang

Department of Atmospheric Sciences, School of Earth Sciences, Zhejiang University, Hangzhou, China

Keywords:Solar radiation modification Geoengineering Carbon cycle Ocean acidification Ocean biogeochemistry

ABSTRACT Solar radiation modification (SRM, also termed as geoengineering) has been proposed as a potential option to counteract anthropogenic warming.The underlying idea of SRM is to reduce the amount of sunlight reaching the atmosphere and surface, thus offsetting some amount of global warming.Here, the authors use an Earth system model to investigate the impact of SRM on the global carbon cycle and ocean biogeochemistry.The authors simulate the temporal evolution of global climate and the carbon cycle from the pre-industrial period to the end of this century under three scenarios: the RCP4.5 CO 2 emission pathway, the RCP8.5 CO 2 emission pathway, and the RCP8.5 CO2 emission pathway with the implementation of SRM to maintain the global mean surface temperature at the level of RCP4.5.The simulations show that SRM, by altering global climate, also affects the global carbon cycle.Compared to the RCP8.5 simulation without SRM, by the year 2100, SRM reduces atmospheric CO 2 by 65 ppm mainly as a result of increased CO 2 uptake by the terrestrial biosphere.However,SRM-induced change in atmospheric CO 2 and climate has a small effect in mitigating ocean acidification.By the year 2100, relative to RCP8.5, SRM causes a decrease in surface ocean hydrogen ion concentration ([H + ]) by 6%and attenuates the seasonal amplitude of [H + ] by about 10%.The simulations also show that SRM has a small effect on globally integrated ocean net primary productivity relative to the high-CO2 simulation without SRM.This study contributes to a comprehensive assessment of the effects of SRM on both the physical climate and the global carbon cycle.

1.Introduction

Since pre-industrial times, human activities have caused an increase in concentrations of greenhouse gases, which is the main contributor to global warming ( IPCC, 2021 ).To limit global warming, various climate intervention approaches have been proposed.Among them, solar radiation modification (SRM, also termed as geoengineering) aims to cool the Earth by deliberately decreasing shortwave radiation reaching the surface to offset some of the warming effects of greenhouse gas emissions ( Crutzen, 2006 ; IPCC, 2021 ).Some of the widely proposed methods of SRM include marine cloud brightening, space shading, stratospheric aerosol injection, and cirrus cloud thinning ( Kalidindi et al.,2015 ; Duan et al., 2018 ; Stjern et al., 2018 ).

To date, most studies have focused on the effect of SRM on physical properties of the climate system such as temperature, precipitation,ocean circulation, and sea ice ( Kravitz et al., 2013 ; Irvine et al., 2016 ;Lawrence et al., 2018; Park et al., 2019 ).However, the effect of SRM on the carbon cycle is much less studied ( Cao, 2018 ).SRM approaches,by altering global climate, could also influence the terrestrial and ocean carbon cycle.Previous studies have shown that stratospheric aerosol injection could increase the terrestrial carbon sink ( Tjiputra et al., 2016 ;Muri et al., 2018 ; Yang et al., 2020 ).

A few studies have focused on the response of the ocean carbon cycle and marine biogeochemistry to SRM ( Tjiputra et al., 2016 ; Cao and Jiang, 2017 ; Lauvset et al., 2017 ; Muri et al., 2018 ; Yang et al., 2020 ).Tjiputra et al.(2016) examined the impact of stratospheric aerosol injection on the ocean carbon cycle under the RCP8.5 CO2emission scenario and found that the ocean absorbed about 10% more carbon under the simulation with stratospheric aerosol injection than the simulation without SRM.Muri et al.(2018) simulated different SRM approaches including stratospheric aerosol injection, marine sky brightening, and cirrus cloud thinning that brought down the radiative forcing from the level of RCP8.5 to RCP4.5.Their results showed that compared to the RCP8.5 scenario without SRM, implementation of SRM enhanced the cumulative ocean carbon sink from 15 to 18 PgC by the end of this century.Some studies have also examined the effect of SRM on ocean acidification ( Tjiputra et al., 2016 ; Cao and Jiang, 2017 ; Lauvset et al.,2017 ), the results of which show that SRM approaches would not alleviate the acidification of the surface ocean.Lauvset et al.(2017) investigated the impacts of SRM on ocean biogeochemistry and ocean net primary productivity (NPP).Their results showed that compared to the RCP4.5 simulation without SRM, stratospheric aerosol injection, marine cloud brightening, and cirrus cloud thinning caused a global decrease of 6%, 6%, and 3% in ocean NPP by the year 2100 (relative to the mean of 1971–2000), respectively.

In this study, we build on previous studies to further investigate the effect of SRM on the ocean carbon cycle and marine biogeochemistry.We use an Earth system model to perform idealized SRM simulations in which solar irradiance is reduced uniformly to bring the global mean temperature from a high-CO2emission scenario to that of a low-CO2emission scenario.Our goal here is to examine the effect of reduced solar radiation on a suite of ocean carbon cycle variables from the seasonal to centennial time scales and the underlying governing processes.Our study contributes to further advancing our understanding of the potential impacts of SRM on the ocean biogeochemical cycles and marine environment.

2.Data and methods

In this study, we used the University of Victoria Earth System Climate Model (ESCM) version 2.10 (UVic ESCM v2.10), which is an Earth system model of intermediate complexity that couples the global climate and the carbon cycle.The model consists of a vertically integrated energy–moisture balance atmospheric model, a land surface and terrestrial carbon cycle model, and a three-dimensional ocean general circulation model ( Weaver et al., 2001 ; Mengis et al., 2020 ).UVic ESCM uses a common horizontal resolution of 3.6° longitude by 1.8°latitude and its ocean module is divided into 19 vertical layers with thickness ranging from 50 m near the surface to 500 m in the deep ocean.

The model was spun up under the pre-industrial CO2concentration of 280 ppm for 10 000 years.The model reached a quasi-equilibrium state for both global climate and the carbon cycle after the 10 000-year simulation (see Text S1 for details).We used the quasi-equilibrium state to represent the nominal year of 1850.From 1850, three transient simulations were performed under prescribed CO2emissions, with all simulations lasting to the year 2100: (1) RCP8.5, in which the model was driven by CO2emissions following the IPCC’s RCP8.5 high-emissions scenario; (2) RCP4.5, in which the model was driven by CO2emissions following the RCP4.5 scenario; and (3) RCP8.5_SRM, in which the model was forced by the same CO2emissions as prescribed in RCP8.5 but with implementation of SRM in such a way that the global mean surface air temperature under RCP8.5 followed that of RCP4.5 starting from the year 2020.

The model-simulated present-day large-scale climate and carbon cycle state agreed well with available observations (Figs.S1–S4).Further details of the model along with relevant references, simulation experiments, and model evaluations are provided in the supplementary material.

3.Results and discussion

3.1. Effect of SRM on carbon storage

By the end of this century, the UVic-simulated global mean warming (relative to the pre-industrial level) for RCP8.5 and RCP4.5 is 3.7°C and 2.1°C, respectively ( Fig.1 (a)).By design, a uniform reduction in solar insolation under RCP8.5 (RCP8.5_SRM) brings the global mean surface air temperature down to that of RCP4.5 ( Fig.1 (a)).However, a substantial difference in the spatial pattern of surface temperature is observed between RCP4.5 and RCP8.5_SRM.Compared to RCP4.5, in RCP8.5_SRM, a slight overcooling is found in the low-to-midlatitude regions and residual warming is found in high-latitude regions(Fig.S5).

By the year 2100, the simulated atmospheric CO2is 538, 927, and 862 ppm for RCP4.5, RCP8.5, and RCP8.5_SRM, respectively ( Fig.1 (b)).Compared to RCP8.5, the reduction in atmospheric CO2in RCP8.5_SRM is a result of SRM-induced cooling that affects carbon uptake by both the terrestrial biosphere and ocean (Fig.S6).

By the year 2100, the total amount of anthropogenic carbon (here,the term anthropogenic carbon refers to change in carbon storage (or flux) relative to the preindustrial state) stored in the terrestrial ecosystem for RCP4.5, RCP8.5, and RCP8.5_SRM is 341, 427, and 571 PgC,respectively ( Fig.2 (a)).In the model, terrestrial carbon uptake is determined by the balance between NPP (which is the difference between gross primary production (GPP) and plant respiration) and soil respiration.Relative to RCP8.5, SRM-induced cooling decreases the terrestrial GPP.The decrease in GPP is mainly observed at high latitudes (Fig.S7),due to cooling-induced reduction in the plant growing season.SRMinduced cooling also decreases vegetation respiration (Fig.S7).The reduction in vegetation respiration dominates the reduction in GPP, and thus compared to RCP8.5, SRM-induced cooling causes an increase in NPP (Fig.S8).Furthermore, SRM-induced cooling suppresses soil respiration (Fig.S8).As a result, the net terrestrial CO2uptake, being the balance between NPP and soil respiration, is greater in RCP8.5_SRM than in RCP8.5 ( Fig.1 (d) and Fig.S8), thus increasing the net terrestrial carbon storge in large parts of land regions ( Fig.2 (b)).Furthermore, relative to RCP8.5, the enhanced CO2uptake by the terrestrial biosphere in RCP8.5_SRM causes a reduction in atmospheric CO2( Fig.1 (b)).

By the year 2100, the total amount of anthropogenic carbon stored by the ocean for RCP4.5, RCP8.5, and RCP8.5_SRM is 393, 553, and 548 PgC, respectively ( Fig.2 (a)).Relative to the high-CO2world without SRM (RCP8.5), SRM affects oceanic CO2uptake in different ways.On the one hand, the SRM-induced decrease in atmospheric CO2tends to reduce ocean CO2uptake as a direct result of the reduction in the air–sea pCO2gradient.On the other hand, SRM-induced cooling tends to increase oceanic CO2uptake.Compared to RCP8.5, a cooler surface ocean in RCP8.5_SRM tends to increase CO2solubility, and the reduction of North Atlantic Deep Water formation (Fig.S9) is weaker, which tends to transport more CO2from the surface to the deep ocean (Fig.S10).In high-latitude ocean areas, the oceanic CO2uptake for RCP8.5_SRM is less than that of RCP8.5 as a result of more sea-ice cover in RCP8.5_SRM,which inhibits CO2uptake (Fig.S10–S11).Overall, in terms of oceanic CO2uptake, the effect of reduced atmospheric CO2and SRM-induced cooling act against each other, resulting in a small change in global ocean carbon storage ( Fig.2 (b)).

Fig.2 (a) shows the allocation of anthropogenic CO2for different scenarios.In summary, by the year 2100, for RCP4.5, 42%, 27%, and 31% of anthropogenic CO2is stored in the atmosphere, land, and ocean,respectively; for RCP8.5, 58%, 18%, and 24% of anthropogenic CO2is stored in the atmosphere, land, and ocean, respectively; and for RCP8.5_SRM, 52%, 24%, and 23% of anthropogenic CO2is stored in the atmosphere, land, and ocean, respectively.

Fig.1.Model-simulated time series of temperature and carbon cycle variables for the scenarios of RCP4.5, RCP8.5, and RCP8.5_SRM: (a) change in global mean surface air temperature (relative to 1850); (b) atmospheric CO 2 concentration; (c) annual CO 2 uptake by the ocean; (d) annual CO 2 uptake by the land; (e) cumulative CO 2 uptake by the ocean; (f) cumulative CO 2 uptake by the land.

3.2. Effect of SRM on ocean acidification

CO2-induced ocean acidification is expected to have many adverse effects on the marine ecosystem ( Bindoffet al., 2019 ).As a result of oceanic uptake of anthropogenic CO2, the global ocean becomes more acidic.By the year 2100, relative to the pre-industrial state, the modelsimulated annual and global mean surface hydrogen ion concentration [H+] increases by 67%, 168%, and 151% for RCP4.5, RCP8.5,and RCP8.5_SRM, respectively ( Fig.3 (a)), corresponding to a decrease in surface ocean pH of 0.22, 0.43, and 0.40 ( Fig.3 (b), Fig.S12).At the same time, the model-simulated reduction in the aragonite saturation state (a less stable form of calcium carbonate mineral) relative to pre-industrial times is 0.91, 1.52, and 1.49 for RCP4.5, RCP8.5, and RCP8.5_SRM, respectively ( Fig.3 (c), Fig.S12).As a reference, our simulated reduction in pH between the period 2090–2099 and 1990–1999 under both RCP4.5 (0.15) and RCP8.5 (0.33) is the same as that of CMIP5 (phase 5 of the Coupled Model Intercomparison Project) ensemble mean results ( Bopp et al., 2013 ) during the same period.

Compared to the high-CO2simulation of RCP8.5, SRM only has a minor effect on ocean acidification.To investigate in detail the mechanisms of how SRM affects ocean acidification, additional offline calculations were performed to quantify the contributions of individual factors to changes in [H+], pH, and the aragonite saturation state.As shown in Fig.S13, the SRM-induced reduction in SST alters ocean acidification in several ways.On the one hand, the SRM-induced reduction in sea surface temperature (SST) directly acts to decrease [H+] and increase pH but decrease the aragonite saturation state.This is because of the different dependence of pH and the aragonite saturation state on temperature ( Cao et al, 2007 ).On the other hand, SRM-induced cooling acts to increase the CO2solubility, resulting in increased dissolved inorganic carbon (DIC) at surface and decreasing both pH and aragonite saturation state.SRM-induced cooling also tends to suppress the rate of calcification, which is parameterized as a function of temperature in the model.Suppressed calcification increases the surface ocean alkalinity,and thus enhances both the pH and aragonite saturation state.Contributions from changes in salinity are rather small.Therefore, compared to RCP8.5, the net effect of SRM is to slightly decrease [H+] and increase the pH, but barely have any effect on aragonite saturation.

Fig.2.(a) Model-simulated allocation of anthropogenic carbon emissions for the scenarios of RCP4.5, RCP8.5, and RCP8.5_SRM for the year 2100.(b) Spatial pattern of the difference in global carbon storage for RCP8.5_SRM relative to RCP8.5.Values were calculated using the simulation results averaged over 2090–2100.

Oceanic uptake of anthropogenic CO2acidifies the deep ocean more slowly than the surface ocean because of the long time scale associated with DIC transport from the surface to the deep ocean.By the year 2100,the global mean whole-ocean pH is almost the same for RCP4.5, RCP8.5,and RCP8.5_SRM.Larger change in deep ocean pH is observed in the North Atlantic Ocean as a result of the Atlantic Meridional Overturning Circulation (AMOC), which brings the DIC from the surface to the deep ocean in a relatively short time scale ( Tjiputra et al., 2010 ).By the year 2100, relative to the pre-industrial period, the deep ocean (depth>2 km) pH in the North Atlantic decreases by 0.19 for RCP8.5.SRM-induced cooling reduces the weakening of the AMOC, leading to an additional pH reduction of 0.01 (relative to RCP8.5) in this region (Fig.S14).

Next, we assess the simulated changes in the seasonality of the ocean carbonate chemistry variables ( Fig.3 (d–f)).The seasonal variation of ocean acidification will have important effects on the functioning of ocean biota and marine ecosystems ( Wang et al., 2017 ).We calculated the seasonality of ocean carbonate chemistry variables following the method of Kwiatkowski and Orr (2018) .We first derived a cubic spline fit from the annual mean time series of modeled carbonate fields, and then we detrended the time series of monthly mean data by subtracting the derived cubic spline fit.Next, we used the detrended monthly mean time series data to determine the seasonal amplitude (difference between the maximum monthly mean value of a carbonate variable and its minimum monthly mean value) of carbonate chemistry variables for every year.

Our simulations show that, by the year 2100, the global mean seasonal amplitude of [H+] is increased by 56%, 117%, and 105% for RCP4.5, RCP8.5, and RCP8.5_SRM, respectively (relative to the preindustrial state).However, the seasonal amplitude of pH attenuates for all the simulations.According to the definition of changes in pH( Kwiatkowski and Orr, 2018 ),

the seasonal amplitude of pH depends on both the seasonal amplitude of [H+] and the annual trend.For example, by the year 2100, under RCP8.5, the seasonal amplitude of [H+] increases by 117%, and the annual mean [H+] increases by 168%, resulting in an 18% reduction in the seasonal amplitude of pH.By the year 2100, the global mean seasonal amplitude of the aragonite saturation state is reduced by 16%, 30%, and 31% for RCP4.5, RCP8.5, and RCP8.5_SRM, respectively ( Fig.3 (f)).

The seasonal amplitude of ocean carbonate fields mainly depends on the seasonal amplitude of DIC, alkalinity and temperature.The mechanism that drives the change in the seasonal amplitude of [H+] and aragonite saturation can be mainly attributed to the changes in buffering capacity of the ocean.As CO2keeps increasing, the ocean’s absorption of CO2reduces the ocean pH, causing a reduced buffering capacity of the ocean ( Egleston et al., 2010 ; Hauck and Voelker, 2015 ).The decreased ocean buffering capacity increases the sensitivity of [H+] change to changes in DIC and thus leads to an amplification of the seasonal amplitude of [H+].As shown in Fig.S15, at constant alkalinity and temperature, for the same amount of DIC change, changes in [H+] would be larger when the DIC concentrations becomes larger (for example, for a DIC concentration change from 2000 to 2100μmol kg−1, [H+] changes from 7.4 to 11.6 nmol kg−1; and for the same range of DIC change from 2200 to 2300μmol kg−1, [H+] changes from 20.9 to 42.8 nmol kg−1).However, for the same amount of DIC change, changes in the aragonite saturate state would be smaller when the DIC concentrations become larger.

Relative to RCP8.5, SRM-induced cooling reduces the seasonal amplitude of [H+] slightly, but has no evident effect on the seasonal amplitude of pH and aragonite saturation.This is because SRM-induced cooling mainly mutes the radiative effect of increasing CO2, which makes a minor contribution to the seasonal amplitude of ocean carbonate chemistry variables ( Kwiatkowski and Orr, 2018 ).

3.3. Effect of SRM on ocean NPP

NPP is an integrated indicator of productivity of the ocean ecosystem and is a key component of the biological pump that regulates the oceanic uptake of CO2.The simulated global ocean NPP averaged over the years 1995 to 2005 is 67.5 PgC yr−1, which is at the high end of the estimated range of 44–67 PgC yr−1from the same period ( Westberry et al, 2008 ).Throughout the 21st century, the model-simulated NPP shows modest change.By the year 2100, relative to the pre-industrial level, the globally integrated NPP decreases by 1.5% under RCP4.5 and increases by 0.7% under RCP8.5.Under RCP8.5_SRM, NPP decreases by 1.8%, similar to that of RCP4.5 ( Fig.4 (a)).As a reference, the UVic-projected change in NPP for the mean of the 2090s (relative to the mean of the 1990s) under RCP 4.5 and RCP8.5 is − 1.2% and 1.7%, respectively.For comparison, CMIP5 model projected changes in NPP during the same period (mean of the 2090s relative to the mean of the 1990s) under RCP 4.5 and RCP8.5 is − 3.6 ± 5.7% and − 8.6 ± 7.9%, respectively ( Bopp et al.,2013 ).CMIP6 (phase 6 of the Coupled Model Intercomparison Project)projected changes in NPP (mean of 2080–2099 relative to the mean of 1870–1899) are − 1.1 ± 5.8% and − 3.0 ± 9.1%, respectively, under the scenarios of SSP2-4.5 and SSP5-8.5 ( Kwiatkowski et al., 2020 ).These results indicate large uncertainty in model-simulated ocean NPP change.

Fig.3.Model-simulated temporal evolution of key carbonate chemistry variables for the scenarios of RCP4.5, RCP8.5, and RCP8.5_SRM: (a–c) the global mean surface ocean (a) [H + ], (b) pH, and (c) aragonite saturation, in which annual mean values are overplotted with monthly mean values; (d–f) the global mean change in the seasonal amplitude of (d) [H + ], (e) pH, and (f) aragonite saturation.

The spatial distributions of projected NPP change are shown in Fig.S16.In some regions, the UVic-simulated pattern of NPP (i.e., RCP8.5)is largely consistent with CMIP5 ensemble mean results ( Bopp et al.,2013 ) (e.g., the tropical Indian and Southern Oceans).In some other regions (e.g., the eastern equatorial Pacific), the UVic-simulated pattern of NPP change is different from that of the CMIP5 ensemble mean,but consistent with some individual CMIP5 models ( Bopp et al., 2013 ).Different changes in ocean temperature, mixing, stratification, or the representation of ocean NPP calculation, might be responsible for these differences.

The change in NPP is governed by a suite of environmental factors including the amount of sunlight, temperature, and the availability of nutrients.To identify the relative contribution of each factor to simulated NPP change, we used monthly mean model output data of solar radiation at the ocean surface, upper-ocean temperature, and nutrient concentrations to diagnose the ocean NPP offline (see Text S4 and Figs.S17–S19 in the supplementary material for details).By allowing only one variable (e.g., temperature) to change with time whilst keeping other variables fixed at pre-industrial levels (e.g., solar radiation and nutrient concentrations), the relative contribution of the change in a certain variable to NPP change can be estimated.

Offline simulations show that increased upper-ocean temperature alone results in a substantial increase in ocean NPP as a result of the warming-induced increase in phytoplankton growth rate.Under RCP8.5, by the year 2100, ocean warming alone increases the global ocean NPP by 13.5%.However, reduced nutrient concentrations in the upper ocean associated with enhanced ocean stratification decrease the ocean NPP by 12.4% ( Fig.4 (b, c)).As a result, the combined effect of changes in temperature and nutrients causes a small increase in globally integrated NPP.Under RCP8.5_SRM, reduced solar radiation brings the upper-ocean temperature and nutrient concentrations close to the level of RCP4.5.Therefore, the relative contributions of temperature and nutrient change to NPP are similar to those of RCP 4.5.The effect of reduced sunlight associated with SRM is small.

Fig.4.Time series of annual and global mean ocean NPP for the scenarios of RCP4.5, RCP8.5, and RCP8.5_SRM.All results are plotted as the percentage change relative to the year 1850.As a reference, the model-simulated ocean NPP in the year 1850 is 67.8 PgC yr − 1 .(a) NPP from model-simulation results.(b, c)NPP change due to individual factors diagnosed from offline calculations.NPP temp represents the NPP change due to temperature changes only, and NPP residual(NPP total − NPP temp − NPP light ) represents the NPP change due to changes in circulation-induced nutrient change.

4.Conclusion

In this study, we used the UVic model to investigate the response of the ocean carbon cycle and ocean acidification to SRM.Our SRM approach was designed in a highly idealized manner that reduced the global mean temperature from the level of RCP8.5 to RCP4.5 by uniformly reducing solar insolation over the globe.Based on the derived relationship between radiative forcing and the sulfate aerosol injection rate ( Niemeier and Timmreck, 2015 ), we estimated the required amount of stratospheric sulfate aerosol injection in the simulation of RCP8.5_SRM (Fig.S20).The maximum sulfate aerosol injection rate needed in the simulation of RCP8.5_SRM is 29 Tg S yr−1.

Our simulations show that compared to the high-CO2world without SRM, reduced solar insolation decreases atmospheric CO2concentrations mainly through enhanced CO2uptake by the terrestrial biosphere.By the year 2100, relative to the scenario of RCP8.5, the implementation of SRM enhances terrestrial CO2storage by 143 PgC and decreases oceanic CO2storage by 5 PgC, resulting in a 65 ppm decrease in atmospheric CO2.In our simulations, SRM has a minor effect on CO2uptake by the ocean as a result of the competing effect of reduced atmospheric CO2that decreases the air–sea CO2gradient and SRM-induced cooling that enhance oceanic carbon uptake.

Our simulations confirm the finding from previous studies( Tjiputra et al., 2016 ; Cao and Jiang, 2017 ; Lauvset et al., 2017 ;Muri et al., 2018 ) that SRM has a minor effect on surface ocean acidification.Here, we further examined the effect of increasing CO2and SRM on the seasonal change of ocean acidification.It was found that increasing atmospheric CO2is the main contributor to the change in the seasonality of ocean acidification.By the year 2100, under RCP8.5, the seasonal amplitude of [H+] increases by 117% and the seasonal amplitude of pH decreases by 18%.Our results are broadly consistent with the study of Kwiatkowski and Orr (2018) , who analyzed nine Earth system model results forced with the RCP8.5 scenario.They reported that by the end of this century, the seasonal amplitude of [H+] would be amplified by 81% ± 16%, and the seasonal amplitude of pH attenuated by 16% ± 7%, respectively.In our simulations, SRM-induced cooling has a modest effect on the seasonality of ocean acidification.By the year 2100, under RCP8.5_SRM, the seasonal amplitude of [H+] increases by 105% and the seasonal amplitude of pH decreases by 18%.

CO2-induced warming also affects ocean NPP mainly by the direct effect of increased SST, which boosts the growth rates of phytoplankton and indirectly affects changes in upper-ocean nutrient concentrations associated with ocean stratification.In our simulations, the effect of increased SST and reduced upper-ocean nutrient availability largely offset each other, resulting in small changes in the globally integrated ocean under scenarios with/without SRM.However, at regional scales, change in NPP is more heterogeneous, indicating different strengths of the temperature effect and ocean mixing effect on NPP in different regions.By the year 2100, relative to RCP8.5, RCP8.5_SRM decreases the global ocean NPP by 2.5% (1.7 PgC).For comparison, Lauvset et al.(2017) simulated three different SRM methods (stratospheric aerosol injection, marine sky brightening, and cirrus cloud thinning) and found that, relative to the RCP4.5 scenario, stratospheric aerosol injection and marine sky brightening cause a reduction in ocean NPP of 6% by the year 2100,and cirrus cloud thinning causes a reduction in ocean NPP of 3%.

There are some limitations and uncertainties in our simulations.For instance, in our simulation, SRM is represented by directly reducing solar insolation.The effect of other SRM approaches on the ocean carbon cycle and biogeochemistry could be different.For example, some studies( Tjiputra et al., 2016 ; Muri et al., 2018 ) show stratospheric aerosol injection enhances the ocean carbon sink compared to high-CO2simulation without SRM, while our results show a small decrease in ocean carbon storage.Oceanic CO2uptake mainly depends on the amount of CO2in the atmosphere and the climate state.As discussed earlier, SRM-induced cooling tends to increase oceanic CO2uptake, but an SRM-induced reduction in atmospheric CO2tends to decrease oceanic CO2uptake.The competing effect of CO2and temperature differs among models, which depends on multiple factors such as the ocean carbon cycle response,the climate response to different SRM approaches and strategies, and the terrestrial carbon cycle response.Existing uncertainties in the ocean carbon cycle model, especially the marine biological component, will affect the model-simulated ocean CO2uptake, ocean NPP, and seasonality of ocean acidification.Also, the current version of the UVic model that we used has no nitrogen cycle processes.Carbon–nitrogen feedback plays a key role in the terrestrial carbon cycle and its response to climate change and SRM ( Thornton et al., 2009 ; Glienke et al., 2015 ; Duan et al.,2020 ).The response of the terrestrial carbon cycle will affect the ocean carbon cycle via impacts on atmospheric CO2.Our results are based on single-model simulations.Further analyses using multiple models are needed to narrow down the uncertainty in the effects of SRM on the ocean carbon cycle and biogeochemistry.

In conclusion, our results demonstrate the potential effects of SRM on the ocean carbon cycle, ocean acidification, and marine NPP, thereby helping towards a comprehensive assessment of the climatic and environmental effects of SRM.

Funding

Long Cao, Xiaoyu Jin, and Jingyu Zhang are supported by the National Natural Science Foundation of China [grant number 41975103 ].

Acknowledgments

All data and UVic results used in the paper are archived at the supercomputer center at Zhejiang University, which can be obtained by contacting the corresponding author (longcao@zju.edu.cn).

Supplementary materials

Supplementary material associated with this article can be found, in the online version, at doi: 10.1016/j.aosl.2022.100187 .