1. Introduction
The subpolar North Atlantic (SPNA; 47°–65°N, 0°–60°W) is a critical region for the global climate system, as ocean deep convection and dense water formation in this region is part of the Atlantic meridional overturning circulation (AMOC), which is crucial for regulating global heat and carbon transport (Buckley and Marshall 2016). The AMOC strength on decadal time scales and longer is mainly affected by SPNA seawater density driven by surface heat and freshwater fluxes (Jackson and Petit 2023). Recent observations (Smeed et al. 2018) and climate model simulations suggest a downward trend of the AMOC strength in recent decades, and modeling studies generally agree on a weakening AMOC in the twenty-first century (Weijer et al. 2020). However, the extent of future AMOC changes can be model dependent, and the processes behind such AMOC changes and their impacts on regional climate are uncertain (Drijfhout and Hazeleger 2007; Cheng et al. 2013; Kostov et al. 2014). Understanding the physical processes behind the different AMOC responses is essential for risk and uncertainty quantifications involved with future climate change.
Research on the sensitivity of the AMOC under future climate conditions primarily has focused on the influence of greenhouse gases and surface warming. While studies have demonstrated that explosive volcanic eruptions and subsequent increases in stratospheric aerosols can influence the North Atlantic climate (Swingedouw et al. 2017), these episodic injections of aerosols from tropical sources offer only a limited perspective on the AMOC’s overall response compared to greenhouse gas–induced changes. To overcome this limitation, an alternative approach involves specifying an idealized global distribution of stratospheric aerosols and studying the responses of the AMOC and tropospheric climate to these aerosols. Previous experiments that simulated increased stratospheric aerosols have indicated a reduction in the AMOC weakening observed in conventional global warming scenarios with elevated greenhouse gas concentrations. This reduction can be attributed to processes associated with differences in surface air–ocean temperatures and the decline of September Arctic sea ice (Xie et al. 2022).
However, there are likely other processes involved with future changes in AMOC that can be elucidated by climate model sensitivity experiments where increased stratospheric aerosols are specified, allowing for a comparative analysis of AMOC responses in contrast to conventional global warming experiments with increasing greenhouse gases. In particular, there may be significant processes associated with the interplay between the hydrological responses of the SNPA, the strength of the AMOC, and the Arctic climate, as the Arctic region plays a significant role in Earth’s climate and is sensitive to various external forcings, including stratospheric aerosols. Previous research has shown that Arctic sea ice changes could influence the AMOC through both the thermal and freshwater effects (Sévellec et al. 2017). On multidecadal time scales, sea ice melting-induced freshening could reach the ocean deep convection sites in the SPNA, leading to the weakening of the AMOC (Sévellec et al. 2017; Liu et al. 2019; Liu and Fedorov 2019; Li et al. 2021).
Here we examine the physical processes involved with the impact of stratospheric aerosols on the SPNA and the AMOC strength in two large-ensemble experiments with two versions of a comprehensive Earth system model. The SPNA responses to increases of stratospheric aerosols are drastically different in the two experiments with different distributions of stratospheric aerosols and background climate—one becomes warm and salty, and the other cool and fresh. The different responses in SPNA are accompanied by their diverging changes in AMOC and Arctic sea ice responses. Both ensemble simulations use the comprehensive Community Earth System Model with the Whole Atmosphere Community Climate Model as its atmospheric component (CESM-WACCM) but differ in model version and simulation design. Fasullo and Richter (2023) identified three main drivers of the overall differences in temperature patterns to increases of stratospheric aerosols between the two sensitivity experiments, namely rapid adjustment of clouds and rainfall to high CO2, the low-frequency dynamical responses in the AMOC, and the differences in forcing. However, the impact of the increased stratospheric aerosols on the contrasting SPNA salinity responses and their connections with the AMOC remains unknown.
The primary objective of this study is to investigate the physical processes involved with the influence of increased stratospheric aerosols on regional responses in the SPNA, explore their potential connections with the Arctic climate, and assess their impact on the AMOC. We emphasize that this study is not an attempt to achieve a full understanding of the intermodel differences in AMOC behaviors; rather, we focus on understanding the impact of increased stratospheric aerosols on the SPNA salinity responses and employing freshwater budget analysis to unveil its connections with the Arctic and AMOC. The structure of the paper is as follows. Section 2 introduces the models and sensitivity experiments. Section 3 summarizes the ensemble-mean climate responses to increases of stratospheric aerosols, including the upper-ocean temperature and salinity changes and the evolution of the AMOC strength. Section 4 investigates the physical processes behind the SPNA hydrological responses, where we use freshwater budget analysis to understand the sources of the SPNA salinity changes, examine the influence of the Arctic sea ice, and investigate the drivers for the AMOC changes. Results are discussed and summarized in section 5.
2. Data and method
a. Models
The two climate model sensitivity experiments use different versions of CESM-WACCM. The first sensitivity experiment is run with CESM version 1 with the Whole Atmosphere Community Climate Model component version 5 (CESM1-WACCM5) (Mills et al. 2017; Tilmes et al. 2018). The model also includes interactive aerosols (MAM3; Liu et al. 2012; Mills et al. 2017), and biogenic emissions (Guenther et al. 2012). WACCM uses a nominal 1° grid with 70 vertical layers reaching up to 140 km (∼10–6 hPa). The configuration has full stratospheric dynamics and interactive middle-atmospheric chemistry. The atmosphere model can explicitly simulate sulfate aerosol concentrations and size distributions.
The second sensitivity experiment uses CESM version 2 with the Whole Atmosphere Community Climate Model version 6 as its atmospheric component (CESM2-WACCM6) (Gettelman et al. 2019; Danabasoglu et al. 2020; Richter et al. 2022). CESM2-WACCM6 was used to contribute climate change projection simulations to phase 6 of the Coupled Model Intercomparison Project (CMIP6) (Eyring et al. 2016). The model uses the same atmosphere horizontal and vertical resolutions as CESM1-WACCM5 but has numerous improvements (see Gettelman et al. 2019), including fully interactive tropospheric chemistry and an interactive crop model.
CESM1-WACCM5 and CESM2-WACCM6 use the same ocean component (POP2; Smith et al. 2010; Danabasoglu et al. 2012) with a nominal 1° horizontal resolution on a displaced-pole grid. The horizontal resolution has a uniform 1.125° grid spacing in the zonal direction and varies from 0.27° in the tropics to 0.64° in the extratropical Northern Hemisphere in the meridional direction. The model has 60 vertical levels down to 6000 m depth. The sea ice component is the Los Alamos sea ice model (CICE version 4; Holland 2013) for CESM1 and CICE version 5 for CESM2. The sea ice model has the same model grids as the ocean.
b. The simulations
The reference simulation for comparison to the first sensitivity experiment uses the RCP8.5 emission scenario and is named CESM1-RCP8.5. The CESM1-RCP8.5 simulation is run from 2005 through 2030 with 17 ensemble members, with an additional 3 members continuing through 2099. In the first increased stratospheric aerosol experiment, sulfur dioxide (SO2) amounts are specified at 23–25 km altitude at 15° and 30° latitude in both hemispheres every year from 2020 to keep the global mean temperature, equator-to-pole, and pole-to-pole temperature gradients near 2020 levels (Tilmes et al. 2018). This experiment is named T-2020. T-2020 runs from 2020 until 2099 with 20 ensemble members.
The reference simulation for the second sensitivity experiment is the moderate SSP2-4.5 future scenario (Burgess et al. 2020), named CESM2-SSP2-4.5. The second sensitivity experiment begins in 2035 and aims to keep the global surface temperatures at ∼1.5°C above the preindustrial level and is called T-1.5 (Richter et al. 2022). Both CESM2-SSP2-4.5 and T-1.5 consist of 10 ensemble members. Similar to T-2020, stratospheric aerosols are injected at 15° and 30° in each hemisphere, but at a lower altitude of about 21 km. Figure 1 shows the yearly stratospheric aerosol mass injections specified in T-2020 and T-1.5. Given the different reference climates, the total stratospheric aerosol mass injections are expectedly greater in T-2020 than in T-1.5. The distribution of injection rates across different latitudes is chosen to also maintain consistent interhemispheric temperature gradients. As a result, the two sensitivity simulations also show different latitudinal distributions of total injections: T-2020 has most of the aerosols injected at 30°N and 30°S, and a modest injection at 15°N, while the aerosol injections in T-1.5 primarily occur at 15°S (Fasullo and Richter 2023).
Time series of yearly sulfur dioxide injections (unit: Tg SO2 yr−1, where Tg ≡ 1012 g) at the four injection latitudes (30°S, 15°S, 15°N, 30°N) for (a) T-2020 and (b) T-1.5.
Citation: Journal of Climate 36, 21; 10.1175/JCLI-D-23-0225.1
3. Climate response to increased stratospheric aerosols
a. Salinity and temperature responses
Figures 2 and 3 show the ensemble annual mean sea surface temperature (SST) and sea surface salinity (SSS) responses in the two sensitivity experiments, respectively. We analyze the changes between the period of 2050–69 and 2020–39 in both the reference simulations and the increased stratospheric aerosol sensitivity experiments. Both T-2020 and T-1.5 succeed at keeping the global mean temperature within the target, but there exist considerable regional disparities in SST and SSS.
Surface temperature differences (unit: °C) between (a) CESM1 RCP8.5 (2050–69) and RCP8.5 (2020–39), (b) CESM1 T-2020 (2050–69) and RCP8.5 (2020–39), (c) CESM2 SSP2-4.5 (2050–69) and SSP2-4.5 (2020–39), and (d) CESM2 T-1.5 (2050–69) and SSP2-4.5 (2020–39).
Citation: Journal of Climate 36, 21; 10.1175/JCLI-D-23-0225.1
Surface salinity differences (unit: g kg−1) between (a) CESM1 RCP8.5 (2050–69) and RCP8.5 (2020–39), (b) CESM1 T-2020 (2050–69) and RCP8.5 (2020–39), (c) CESM2 SSP2-4.5 (2050–69) and SSP2-4.5 (2020–39), and (d) CESM2 T-1.5 (2050–69) and SSP2-4.5 (2020–39).
Citation: Journal of Climate 36, 21; 10.1175/JCLI-D-23-0225.1
The SPNA experiences the most dramatic changes in the period of 2050–69 relative to the period of 2020–39. In CESM1-RCP8.5 (Fig. 2a), global SST generally increases, with the most significant warming occurring in the Northern Hemisphere mid-to-high latitudes. In particular, the warming over SPNA exceeds 2°C for the latter period relative to the earlier period. In T-2020 (Fig. 2b), the global SST is kept stable, but the SPNA still shows pronounced regional warming. Conversely, in both the CESM2-SSP2-4.5 and the T-1.5 simulations, despite the global SST increase, a “cold blob” (also known as the “warming hole”) emerges in the SPNA (Figs. 2c,d)—a feature indicating a weakened AMOC and the reduced northward heat transport (Liu et al. 2020). With increased stratospheric aerosols, the temperature of the cold blob is cooler in the T-1.5 simulation than in the CESM2-SSP2-4.5.
The salinity responses in the two models also exhibit contrasting patterns, as shown in Fig. 3. In CESM1-RCP8.5 (Fig. 3a), the subtropical North Atlantic, the SPNA, and the Nordic seas show enhanced salinification, while the Arctic and the tropical oceans experience freshening. In T-2020 (Fig. 3b), there is a dipole pattern in the North Atlantic, with the subtropical and midlatitudinal regions becoming fresher, while the subpolar regions become more saline that extends into the Arctic Ocean. On the other hand, the CESM2-SSP2-4.5 and the T-1.5 simulations show similar patterns of salinity responses featuring a pronounced freshening over SPNA with a bit smaller magnitude for the latter (Figs. 3c,d).
To further examine the different SPNA responses in salinity and temperature among the two ensemble simulations, we analyze the ensemble-annual mean upper-ocean temperature and salinity vertical profiles over this region (Fig. 4). Compared with the base state during 2020–39, CESM1-RCP8.5 and T-2020 during 2050–69 both show increased salinity and temperature across the upper layer, consistent with the surface responses. CESM1-RCP8.5 warms more but becomes less saline than T-2020. As a result, the SPNA upper-ocean seawater density decreases in the CESM1-RCP8.5 experiment but increases in T-2020. Thus, both salinity and temperature contribute the same way to the changes of the upper-ocean density in these two simulations.
Vertical profiles of (left) salinity (g kg−1), (center) temperature (°C), and (right) density (g cm−3) in (top) CESM1-RCP8.5 and T-2020 and (bottom) CESM2-SSP2-4.5 and T-1.5. The black curves are the respective mean during 2020–39; the red curves are the mean of 2050–69 in RCP8.5/SSP2-4.5; the blue curves are the mean of 2050–69 in T-2020 and T-1.5.
Citation: Journal of Climate 36, 21; 10.1175/JCLI-D-23-0225.1
The SPNA in both CESM2-SSP2-4.5 and T-1.5 exhibits almost identical freshening across the upper layer. The SST “cold blob” in CESM2-SSP2-4.5 is relatively shallow, extending down to about 200 m. The temperature in T-1.5 is around 0.7°C cooler than CESM2-SSP2-4.5. Although the upper-ocean density is reduced in both simulations due primarily to the freshening, the cooler upper ocean in T-1.5 makes the upper ocean denser than that in CESM2-SSP2-4.5.
Note that the two ensemble simulations have notable differences in the base state during 2020–39 (Fig. 4, black curves). CESM2-SSP2-4.5 has lower surface salinity (about 0.5 g kg−1 lower) and a greater vertical salinity gradient than CESM1-RCP8.5, which corresponds to stronger stratification and thus is less susceptible to vertical mixing and deep convection. The T-1.5 simulations during 2050–69 further strengthen the stratification, which makes it more difficult to sustain the deep convection in SPNA. These differences are likely intrinsic to the model, and the associated processes and feedbacks may play an important role in their different AMOC responses.
These results show that the SPNA responses differ considerably among the simulations. Both CESM1-RCP8.5 and T-2020 forcings lead to a warmer and saltier SPNA in 2050–69 compared to that in 2020–39. In contrast, the CESM2-SSP2-4.5 and T-1.5 simulations show a cooler and fresher SPNA in 2050–69 relative to 2020–39. Particularly, the significant differences in the SPNA SST responses between CESM1-RCP8.5 and CESM2-SSP2-4.5 suggest that the two models have intrinsic differences in how they respond to greenhouse gas–induced warming.
Fasullo and Richter (2023) found that the different warming patterns among the models can be predominantly attributed to the differences in the rapid adjustments of clouds and absorbed short-wave radiation to elevated levels of CO2. They demonstrated a strong spatial correlation between the intermodel difference in the net absorbed top-of-atmosphere shortwave radiation and the contrasting warming patterns. Specifically, in the CESM2 model (in CESM2-SSP2-4.5 and T-1.5), a significantly smaller net shortwave radiation was observed over the SPNA compared to CESM1 (in CESM1-RCP8.5 and T-2020) (Fig. 2 in Fasullo and Richter 2023). These disparities stem from intrinsic differences in their atmosphere model physics (WACCM5 vs WACCM6). In addition, the study indicated that the future background scenario also plays a role in the intermodel differences, though its contribution is not dominant. Furthermore, the contrasting SPNA responses may be partially attributed to the AMOC responses in the two models. In the CESM1 model, a stronger AMOC transports more heat and salt northward, leading to a warmer and saltier SPNA. Conversely, the CESM2 model exhibits a weaker AMOC, resulting in a cooler and fresher SPNA. The intricate two-way relationship between the SPNA and AMOC presents a challenge in untangling their causal connection.
These intermodel differences suggest that, even with similar approaches to increase stratospheric aerosols that could counteract a rise in global SST, the response of different models to these increased stratospheric aerosols can vary significantly. This highlights the complexity of regional responses to varying climate forcings and underscores the influence of inherent model differences.
b. AMOC and freshwater content responses
Figure 5a compares the time series of the AMOC strength, which is defined as the meridional streamfunction maxima below 500 m depth in the North Atlantic. At the beginning of the simulation, T-2020 and T-1.5 have comparable AMOC strength [23 Sv (1 Sv ≡ 106 m3 s−1) in T-2020 and 22 Sv in T-1.5]. The CESM1-RCP8.5 and the CESM2-SSP2-4.5 reference simulations both show a weakening AMOC over time: the AMOC in CESM1-RCP8.5 slightly increases in the first 15 years before turning to a weakening trend around the year 2035; the AMOC in CESM2-SSP2.4-5 shows a steady and continuous weakening over the course of the simulation. By the end of 2070, the AMOC in CESM1-RCP8.5 decreases by ∼4 Sv, while the AMOC in the CESM2-SSP2-4.5 decreases by 8 Sv. This result is somewhat unexpected because the AMOC experiences less weakening under stronger greenhouse gas forcing, suggesting that the AMOC in CESM2-SSP2-4.5 is more prone to weakening under the projected greenhouse gas forcing. This may be related to the different climate sensitivity between CESM1 and CESM2, as CESM2 has been shown to have a much higher climate sensitivity (5.1°–5.3°C) than CESM1 (4°C) (Danabasoglu et al. 2020; Zhu et al. 2021). Studies by Weijer et al. (2020) and Lin et al. (2019) suggest a potential relationship between the equilibrium climate sensitivity and the mean AMOC strength, with a lower equilibrium climate sensitivity corresponding to a stronger AMOC. This indicates that the AMOC may play a role in determining the overall climate response to increased atmospheric CO2 concentration. Furthermore, Hu et al. (2020) compared CESM2 and the Energy Exascale Earth System Model version 1 (E3SM1)—two climate models with very similar equilibrium climate sensitivity—and demonstrated that a weaker AMOC contributes to a higher transient climate response, as it allows for faster warming of the upper ocean and slower warming of the subsurface ocean.
(top) Time series of AMOC strength (unit: Sv) in CESM1-RCP8.5 (red), T-2020 (orange), CESM2-SSP2-4.5 (green), and T-1.5 (blue) simulations. (bottom) Time series of SPNA upper-ocean (0–300 m) FWC (unit: km3) in the respective simulations.
Citation: Journal of Climate 36, 21; 10.1175/JCLI-D-23-0225.1
In the sensitivity experiments, the AMOC strength in T-2020 steadily increases, opposing the weakening trend seen in the CESM1-RCP-8.5. The AMOC increases by 5 Sv by the end of the simulation. Compared to CESM1-RCP8.5, T-2020 leads to an AMOC change of 9.5 Sv by 2070. Meanwhile, the T-1.5 has a much smaller impact on the AMOC, with only a 2 Sv increase of AMOC strength compared to CESM2-SSP2-4.5 by 2070. Note that, even with the increased aerosol forcing, the AMOC in T-1.5 still weakens more than that in CESM1-RCP8.5.
To better compare the effect of increased stratosphere aerosols on the AMOC strength, we scale the stratospheric aerosol-induced AMOC changes with the amount of effective radiative forcing in the respective simulations. The effective radiative forcing of RCP8.5 is roughly 1.4 times the amount of SSP2-4.5 (IPCC 2021, p. 169). Meanwhile, the global total aerosol injection in T-2020 is roughly twice the amount in T-1.5 (Fasullo and Richter 2023). The AMOC change by 2070 scaled by radiative forcing in T-2020 is therefore 9.5 Sv/1.4/2 = 3.4 Sv, which is still 1.7 times that of T-1.5 (2 Sv). In other words, the increases of stratospheric aerosols in T-2020 are roughly 1.7 times more efficient in restoring AMOC strength than T-1.5. It is possible that the different spatial distributions of the injected aerosols in the two simulations could impact the AMOC responses (Zhang et al. 2023).
The above-mentioned freshwater changes are in line with the anticipated relationship with AMOC. Reduced FWC leads to higher seawater density and subsequently a stronger AMOC. Conversely, a stronger AMOC transports more salt into the SPNA, further reducing the FWC there. This intricate two-way relationship makes it challenging to disentangle their causal connection. In the next section, we will examine the sources of the different freshwater changes in the two sensitivity experiments. We will focus on examining the effect of the increases in stratospheric aerosols by comparing the reference simulations and the respective sensitivity experiments.
4. Mechanisms for the different SPNA hydrological responses
a. Freshwater budget analysis of the SPNA
In this section, we examine the mechanisms behind the SPNA freshwater changes in the model experiments, focusing on understanding whether these changes originate from the AMOC variations or other sources. We compare the freshwater budget in the reference simulations (CESM1-RCP8.5 and CESM2-SSP2-4.5; Figs. 6a,b) and the respective changes induced by increased stratospheric aerosols (Figs. 6c–f), to isolate the effects of those increases in stratospheric aerosols from the different reference climates. We analyze the evolution of the SPNA upper-ocean (0–300 m) freshwater tendency and the contributions from the surface freshwater fluxes in contrast to the ocean processes. The surface freshwater fluxes include precipitation minus evaporation (P − E), sea ice melting and brine rejection, and river runoff; the ocean processes mainly include horizontal and vertical advection, diffusion, and diapycnal mixing.
(a),(b) Time series of SPNA upper-ocean (0–300 m) freshwater tendency (Sv; black) and contributions from surface freshwater fluxes (Sv; red) and ocean processes (Sv; blue) in (a) CESM1-RCP8.5 and (b) CESM2-SSP2-4.5 (Sv). (c),(d) Differences of freshwater tendency between the sensitivity simulation and reference simulations: (c) T-2020 minus CESM1-RCP8.5 and (d) T-1.5 minus CESM2-SSP2-4.5. The ocean processes are further decomposed into vertical (green) and horizontal (orange) processes. (e),(f) Horizontal freshwater convergence anomalies (thick orange) and the contributions from the southern boundary (northward freshwater transport; red) and from the northern boundary (southward freshwater transport; blue). Positive values mean the transport contributes to freshwater convergence.
Citation: Journal of Climate 36, 21; 10.1175/JCLI-D-23-0225.1
In the reference cases, the surface freshwater input and the ocean export generally balance each other. In CESM1-RCP8.5, the total freshwater tendency fluctuates and is relatively stable. CESM2-SSP2-4.5, however, has an overall positive tendency to accumulate freshwater. The positive freshwater tendency in CESM2-SSP2-4.5 comes from the higher surface freshwater flux, including from both P − E and sea ice melting and brine rejection (not shown).
Figures 6c–f show the changes induced by the increases in stratospheric aerosols in the respective simulations. Compared to CESM1-RCP8.5, the total freshwater tendency in T-2020 decreases, with strong decadal variability during the later decades. The decrease is mainly attributed to the ocean processes, where changes in vertical advection and diapycnal mixing dominate, likely resulting from a stronger AMOC that accelerates the transport of freshwater downward (Figs. 6c and 7). The changes in the vertical ocean processes are concomitantly compensated by the horizontal advection that converges freshwater. Meanwhile, there is an increase of surface freshwater input during the early stage of the simulation, which diminishes over time and turns to negative by 2060. We further analyze the sources of horizontal freshwater advection by examining the freshwater transport across the southern and the northern boundaries (Fig. 6e). From 2045, the total horizontal convergence consists of increased freshwater convergence across the southern boundary (increased northward freshwater net transport across the southern boundary) and reduced freshwater convergence from the northern boundary, which indicates reduced freshwater transported from the Arctic to the SPNA. The total change of freshwater tendency and contributions from each term averaged over 2050–69 is summarized in Fig. 7. The negative freshwater tendency in T-2020 is mainly attributed to increased freshwater export resulting from stronger vertical advection and reduced freshwater import from the Arctic (Fig. 7).
Summary of SPNA upper-ocean freshwater tendency differences between SAI and the reference simulations in (left) T-2020 minus CESM1-RCP8.5 and (right) T-1.5 minus CESM2-SSP2-4.5, averaged over 2050–69 (Sv). Note that in T-2020, ocean vertical processes and horizontal advection from the Arctic are the main drivers of the negative total freshwater tendency; in T-1.5, the horizontal advection from the Arctic is the dominant factor. The decreasing FWC is not driven by the AMOC-related northward transport of salty Atlantic water. Rather, the net transport across the southern boundary of SPNA leads to freshwater convergence. In T-2020, the enhanced vertical convection and mixing (due to strengthened AMOC) is important for the reduced upper SPNA FWC. The reduced freshwater export from the Arctic to the SPNA is important for both T-2020 and T-1.5.
Citation: Journal of Climate 36, 21; 10.1175/JCLI-D-23-0225.1
For T-1.5, the FWC tendency changes are much more subtle (Figs. 6d,f). The total tendency turns negative from 2045 and slowly decreases by 0.01 Sv toward 2070. Like T-2020, the decrease is mainly due to ocean processes, and is partly compensated by the increased surface freshwater flux. Contrary to T-2020, the total tendency decrease is mainly due to the horizontal advection of freshwater out of the SPNA (Figs. 6d and 7), which is controlled by the reduced freshwater import from the Arctic.
To summarize, in both sensitivity experiments, the increased stratospheric aerosols tend to reduce the upper SPNA FWC. In T-2020, the enhanced vertical convection and mixing induced by the strengthened AMOC is an important factor for the freshwater content to decrease. However, the effect from the same processes is negligible in T-1.5. Interestingly, a key driver for the reduced upper SPNA FWC in both sensitivity experiments is the reduced freshwater export from the Arctic to the SPNA. In the next section, we will examine the causes of the reduced Arctic freshwater export.
b. Influence of the Arctic and the role of sea ice
(a),(b) Time series of Arctic total freshwater budget (black) and contributions from Bering Strait transport (orange), SPNA transport (green), and surface freshwater fluxes (purple) in (a) CESM1-RCP8.5 and (b) CESM2-SSP2-4.5 (Sv). (c),(d) Differences of Arctic freshwater tendency between the sensitivity experiments and reference simulations: (c) T-2020 minus CESM1-RCP8.5 and (d) T-1.5 minus CESM2-SSP2-4.5 (Sv). Note that the export of freshwater from the Arctic to SPNA decrease in the sensitivity experiments to balance the loss of surface freshwater flux. (e),(f) Total surface freshwater anomalies (purple) and contributions from sea ice melting and brine rejection (blue), river runoff (red), and P − E (orange) (Sv).
Citation: Journal of Climate 36, 21; 10.1175/JCLI-D-23-0225.1
Figures 8c and 8d show the respective changes induced by increased stratospheric aerosols. In T-2020, the total freshwater tendency is on average reduced by 0.025 Sv by 2070, primarily driven by the decrease of surface fluxes (which reduces by 0.075 Sv). In response to the surface freshwater loss, the SPNA freshwater convergence significantly increases. The Bering Strait also contributes to the freshwater convergence, but the change is much smaller. Similar processes are also found in T-1.5, though the magnitude of changes is small—the surface freshwater decrease is about 0.016 Sv, accounting for roughly 1/5 of that in T-2020.
What causes the Arctic surface freshwater loss? Figures 8e and 8f show the contributions from precipitation minus evaporation (P − E), sea ice, and river runoff to the total surface freshwater flux changes. In both T-2020 and T-1.5, freshwater fluxes from all three terms decrease. In T-2020, sea ice processes dominate the total decrease until 2050. It then slightly increases and stabilizes to around −0.017 Sv. Meanwhile, P − E and river runoff keep decreasing, and P − E becomes the dominant factor over the later decades. In T-1.5, each term contributes roughly equally to the total freshwater decrease. The total surface freshwater flux tends to stabilize toward the end of the simulation.
The influence of Arctic sea ice is further illustrated in Fig. 9. The sea ice concentration is smaller in CESM1-WACCM5 than that in CESM2-WACCM6 in the base climate during 2020–39. The larger sea ice extent over the SPNA in CESM2-WACCM6 may be related to its upper-ocean freshness and relatively cooler ocean (Fig. 4). In the reference simulations, the sea ice volume decreases by more than 90% by the end of 2070 in CESM1 RCP8.5 and ∼30% in CESM2 SSP2-4.5. In both cases, the implementation of increased stratospheric aerosols prevents such sea ice decline and keeps a stable amount of sea ice until the end of the simulation. Therefore, the increased stratospheric aerosols reduce the freshwater input resulting from sea ice melting, shown as reduced surface freshwater flux in Figs. 8e and 8f.
Annual mean sea ice concentration (unit: %) in (a) CESM1 RCP8.5 (2020–39), (b) CESM1 RCP8.5 (2050–69), (c) CESM1 T-2020 (2050–69), and (d) the difference between (c) and (b). Sea ice concentration in (e) CESM2 SSP2-4.5 (2020–39), (f) CESM2 SSP2-4.5 (2050–69), and (g) CESM2 T-1.5, and (h) the difference between (g) and (f). (i) Time series of Arctic sea ice volume (unit: km3) in CESM1 T-2020 (orange), CESM1 RCP8.5 (red), CESM2 T-1.5 (blue), and CESM2 SSP2-4.5 (green).
Citation: Journal of Climate 36, 21; 10.1175/JCLI-D-23-0225.1
Therefore, compared with the reference simulations, the Arctic Ocean loses freshwater from the surface in both T-2020 and T-1.5. The reduced surface freshwater input is partly due to changes in P − E and river runoff, and partly stored in the form of sea ice. To compensate for the surface freshwater loss, freshwater export from the Arctic to SPNA decreases, which leads to the decline of SPNA FWC in both cases. Therefore, the connection with the Arctic is key to the SPNA hydrological responses.
5. What causes the diverging AMOC responses?
Section 3b showed that the AMOC responds quite differently to the increased stratospheric aerosols in the sensitivity experiments: The AMOC strength in T-2020 drastically increases, reversing the weakening trend in the CESM1-RCP-8.5 experiment, whereas the AMOC in T-1.5 becomes stronger relative to CESM2-SSP2-4.5, but the weakening trend remains. How do the different SPNA hydrological responses relate to the diverging AMOC behaviors in the two sensitivity simulations? How does the reduced SPNA upper-ocean freshwater contribute to AMOC strengthening?
To understand the physical processes and mechanisms behind the AMOC changes, we analyze the surface density flux anomalies and examine the contributions from surface heat and freshwater fluxes. Figure 10 shows the surface density flux anomalies in the sensitivity experiments (differences between the sensitivity experiments and the corresponding reference simulations) averaged over 2050–69. In both cases, surface density flux anomalies are directly related to March mixed layer depth anomalies, confirming that surface density flux greatly influences the strength of deep convection and therefore the AMOC strength.
Surface density flux anomalies (unit: kg m−2 s−1) induced by surface (a),(e) heat flux, (b),(f) freshwater flux, and (c),(g) their combined effect in (a)–(c) T-2020 and (e)–(g) T-1.5. (d),(h) March ocean mixed layer depth anomalies (unit: m) in (d) T-2020 and (h) T-1.5. The anomalies are referenced to their respective reference simulations averaged over 2050–69.
Citation: Journal of Climate 36, 21; 10.1175/JCLI-D-23-0225.1
In T-2020, there is a strong increase of surface density flux over the Labrador Sea and the Irminger Sea, indicating enhanced deep convection as a result of increased stratospheric aerosols. The increased density flux is primarily dominated by surface heat fluxes, which may be related to changes in radiative heating, increase of sea ice cover, and surface cooling associated with thin ice growth. The changes in surface heat flux (2050–69 compared to 2020–39) in the North Atlantic (Fig. S1 in the online supplemental material) could be linked to the SST anomalies via a negative turbulent heat flux feedback, also known as the North Atlantic redistribution feedback (i.e., Liu and Fedorov 2019; Couldrey et al. 2021), which has been suggested to play an important role in the AMOC. We find that this feedback may be in effect in the CESM2-SSP2-4.5 and T-1.5 simulations, where an SPNA cold blob is present. The cold blob leads to an increased heat uptake into the ocean, acting as negative feedback to the cooling. This may in part contribute to the reduced deep convection and the subsequent AMOC weakening. Because of the simulation design that chose the distribution of aerosol injection at different latitudes to maintain the interhemispheric temperature gradient, there is also potentially a positive feedback wherein the cold blob in CESM2 simulations preferentially reduces Northern Hemisphere temperatures, leading to relatively less injection of aerosols in the Northern Hemisphere in T-1.5 as compared with T-2020; this is noted in Fasullo and Richter (2023) but is likely a relatively small effect.
The surface freshwater fluxes in T-2020 contribute to the increased density flux over the subpolar gyre and reduce the surface density flux around the sea ice edges (Fig. 9) due to stronger seasonal sea ice melting. In T-1.5, the total density flux changes are generally smaller. The ocean deep convection sites are around the south of Iceland and the Norwegian Sea. Similar to T-2020, the increased density flux over the deep convection area is primarily due to stratospheric aerosol-induced surface heat flux anomalies.
Therefore, in both cases, the changes in surface total density fluxes are primarily dominated by the changes in surface heat fluxes. The hydrological responses, though acting to reduce the upper-ocean freshwater and increase density, only play a small part.
6. Discussion and conclusions
This study investigates the physical processes involved with the impacts of increased stratospheric aerosols on the SPNA and the strength of the AMOC in two climate model large-ensemble sensitivity experiments. There are substantial differences in SPNA responses among the simulations, which are fundamentally attributed to intermodel differences in rapid adjustments to forcing and future background scenarios (Fasullo and Richter 2023). The diverging model responses highlight that despite implementing similar approaches to increase stratospheric aerosols, different models exhibit significant variations in the physical processes involved with their responses and consequent impacts.
The two models used in these simulations exhibit very different AMOC behaviors, including the average strengths in the base-state climate of 2020–39, and their sensitivity to surface buoyancy forcings. While a full understanding of the intermodel differences in AMOC behavior is beyond the scope of the current study, we emphasize the physical significance of increased stratospheric aerosols on the SPNA and their potential connections with the AMOC.
We find that differences in the base-state AMOC are likely linked to the intermodel variations in the upper-ocean stratification over the SPNA region. Specifically, the CESM1-RCP8.5 has weaker upper-ocean stratification, making it more susceptible to forcing-induced surface buoyancy changes and the subsequent AMOC responses. The smaller Arctic sea ice cover in CESM1-WACCM5 in the base climate may contribute to its sensitivity to surface forcings. Meanwhile, the larger sea ice extent over SPNA in CESM2-WACCM6 could help explain the freshness of the upper ocean and the stronger stratification (Figs. 4 and 9).
In each set of experiments, we compare the sensitivity experiments with increased stratospheric aerosols to the reference climate change simulations to isolate the physical processes that affect AMOC compared to the reference climate. By analyzing the upper-ocean freshwater budget, we find that increased stratospheric aerosols reduce the SPNA FWC and increase the upper-ocean salinity. In both cases, we find a close connection between the SPNA and the Arctic—the SPNA freshwater loss is mainly driven by decreased freshwater export from the Arctic, where there is a total decrease of surface freshwater fluxes resulting from changes in P − E, river runoff, and sea ice growth. We examine how the SPNA salinification relates to the AMOC responses by analyzing the surface density fluxes over the SPNA ocean deep convection sites. We find that the anomalous deep convection induced by elevated stratospheric aerosols is mainly attributed to changes in surface heat fluxes, while the freshwater-induced buoyancy fluxes only play a small role.
The response in the sensitivity experiments with increased stratospheric aerosols elucidates that the regional hydrological processes in the SPNA, which affect AMOC, are largely dependent on the background climate and future greenhouse gas forcings. While the current study may not fully document all the complexities among different models, it provides valuable physical insights into the specific impacts of increased stratospheric aerosols on SPNA salinity changes and their potential connections with the AMOC and the Arctic. These findings shed light on the physical processes that dictate the important connections between the SPNA, the Arctic, the AMOC, and their subsequent feedbacks to the climate system.
Acknowledgments.
Portions of this study were supported by the Regional and Global Model Analysis (RGMA) component of the Earth and Environmental System Modeling Program of the U.S. Department of Energy’s Office of Biological and Environmental Research (BER) under Award DE-SC0022070. This work also was supported by the National Center for Atmospheric Research, which is a major facility sponsored by the National Science Foundation (NSF) under Cooperative Agreement 1852977.
Data availability statement.
All T-2020 and CESM1-RCP8.5 simulations are available to the community via the Earth System Grid, where the simulations are referred to as the “GLENS” experiment (see information at https://www.cesm.ucar.edu/projects/community-projects/GLENS/). Outputs from the CESM2-SSP2-4.5 and T-1.5 simulations (referred to as the “ARISE-SAI-1.5”) are freely available the NCAR Climate Data Gateway at https://doi.org/10.26024/0cs0-ev98 and https://doi.org/10.5065/9kcn-9y79, respectively.
REFERENCES
Buckley, M. W., and J. Marshall, 2016: Observations, inferences, and mechanisms of the Atlantic meridional overturning circulation: A review. Rev. Geophys., 54, 5–63, https://doi.org/10.1002/2015RG000493.
Burgess, M. G., J. Ritchie, J. Shapland, and R. Pielke Jr., 2020: IPCC baseline scenarios have over-projected CO2 emissions and economic growth. Environ. Res. Lett., 16, 014016, https://doi.org/10.1088/1748-9326/abcdd2.
Cheng, W., J. C. H. Chiang, and D. Zhang, 2013: Atlantic meridional overturning circulation (AMOC) in CMIP5 models: RCP and historical simulations. J. Climate, 26, 7187–7197, https://doi.org/10.1175/JCLI-D-12-00496.1.
Couldrey, M. P., and Coauthors, 2021: What causes the spread of model projections of ocean dynamic sea-level change in response to greenhouse gas forcing? Climate Dyn., 56, 155–187, https://doi.org/10.1007/s00382-020-05471-4.
Danabasoglu, G., S. C. Bates, B. P. Briegleb, S. R. Jayne, M. Jochum, W. G. Large, S. Peacock, and S. G. Yeager, 2012: The CCSM4 ocean component. J. Climate, 25, 1361–1389, https://doi.org/10.1175/JCLI-D-11-00091.1.
Danabasoglu, G., and Coauthors, 2020: The Community Earth System Model version 2 (CESM2). J. Adv. Model. Earth Syst., 12, e2019MS001916, https://doi.org/10.1029/2019MS001916.
Drijfhout, S. S., and W. Hazeleger, 2007: Detecting Atlantic MOC changes in an ensemble of climate change simulations. J. Climate, 20, 1571–1582, https://doi.org/10.1175/JCLI4104.1.
Eyring, V., S. Bony, G. A. Meehl, C. A. Senior, B. Stevens, R. J. Stouffer, and K. E. Taylor, 2016: Overview of the Coupled Model Intercomparison Project Phase 6 (CMIP6) experimental design and organization. Geosci. Model Dev., 9, 1937–1958, https://doi.org/10.5194/gmd-9-1937-2016.
Fasullo, J. T., and J. H. Richter, 2023: Dependence of strategic solar climate intervention on background scenario and model physics. Atmos. Chem. Phys., 23, 163–182, https://doi.org/10.5194/acp-23-163-2023.
Gettelman, A., and Coauthors, 2019: The Whole Atmosphere Community Climate Model version 6 (WACCM6). J. Geophys. Res. Atmos., 124, 12 380–12 403, https://doi.org/10.1029/2019JD030943.
Guenther, A. B., X. Jiang, C. L. Heald, T. Sakulyanontvittaya, T. Duhl, L. K. Emmons, and X. Wang, 2012: The Model of Emissions of Gases and Aerosols from Nature version 2.1 (MEGAN2.1): An extended and updated framework for modeling biogenic emissions. Geosci. Model Dev., 5, 1471–1492, https://doi.org/10.5194/gmd-5-1471-2012, 2012.
Holland, M., 2013: The great sea-ice dwindle. Nat. Geosci., 6, 10–11, https://doi.org/10.1038/ngeo1681.
Hu, A., L. Van Roekel, W. Weijer, O. A. Garuba, W. Cheng, and B. T. Nadiga, 2020: Role of AMOC in transient climate response to greenhouse gas forcing in two coupled models. J. Climate, 33, 5845–5859, https://doi.org/10.1175/JCLI-D-19-1027.1.
IPCC, 2021: Climate Change 2021: The Physical Science Basis. V. Masson-Delmotte et al., Eds., Cambridge University Press, 3949 pp.
Jackson, L. C., and T. Petit, 2023: North Atlantic overturning and water mass transformation in CMIP6 models. Climate Dyn., 60, 2871–2891, https://doi.org/10.1007/s00382-022-06448-1.
Kostov, Y., K. C. Armour, and J. Marshall, 2014: Impact of the Atlantic meridional overturning circulation on ocean heat storage and transient climate change. Geophys. Res. Lett., 41, 2108–2116, https://doi.org/10.1002/2013GL058998.
Li, H., A. Fedorov, and W. Liu, 2021: AMOC stability and diverging response to Arctic sea ice decline in two climate models. J. Climate, 34, 5443–5460, https://doi.org/10.1175/JCLI-D-20-0572.1.
Lin, Y.-J., Y.-T. Hwang, P. Ceppi, and J. M. Gregory, 2019: Uncertainty in the evolution of climate feedback traced to the strength of the Atlantic meridional overturning circulation. Geophys. Res. Lett., 46, 12 331–12 339, https://doi.org/10.1029/2019GL083084.
Liu, W., and A. V. Fedorov, 2019: Global impacts of Arctic sea ice loss mediated by the Atlantic meridional overturning circulation. Geophys. Res. Lett., 46, 944–952, https://doi.org/10.1029/2018GL080602.
Liu, W., A. V. Fedorov, and F. Sévellec, 2019: The mechanisms of the Atlantic meridional overturning circulation slowdown induced by Arctic sea ice decline. J. Climate, 32, 977–996, https://doi.org/10.1175/JCLI-D-18-0231.1.
Liu, W., A. V. Fedorov, S.-P. Xie, and S. Hu, 2020: Climate impacts of a weakened Atlantic meridional overturning circulation in a warming climate. Sci. Adv., 6, eaaz4876, https://doi.org/10.1126/sciadv.aaz4876.
Liu, X., and Coauthors, 2012: Toward a minimal representation of aerosols in climate models: Description and evaluation in the Community Atmosphere Model CAM5. Geosci. Model Dev., 5, 709–739, https://doi.org/10.5194/gmd-5-709-2012.
Mills, M. J., and Coauthors, 2017: Radiative and chemical response to interactive stratospheric sulfate aerosols in fully coupled CESM1(WACCM). J. Geophys. Res. Atmos., 122, 13 061–13 078, https://doi.org/10.1002/2017JD027006.
Richter, J., and Coauthors, 2022: Assessing responses and impacts of solar climate intervention on the Earth system with stratospheric aerosol injection (ARISE-SAI): Protocol and initial results from the first simulations. Geosci. Model Dev., 15, 8221–8243, https://doi.org/10.5194/gmd-15-8221-2022.
Sévellec, F., A. V. Fedorov, and W. Liu, 2017: Arctic sea-ice decline weakens the Atlantic meridional overturning circulation. Nat. Climate Change, 7, 604–610, https://doi.org/10.1038/nclimate3353.
Smeed, D. A., and Coauthors, 2018: The North Atlantic Ocean is in a state of reduced overturning. Geophys. Res. Lett., 45, 1527–1533, https://doi.org/10.1002/2017GL076350.
Smith, R. D., and Coauthors, 2010: Parallel Ocean Program (POP) reference manual, Ocean component of the Community Climate System Model (CCSM). Tech. Rep. LAUR-10-01853, Los Alamos National Laboratory, 141 pp., https://www2.cesm.ucar.edu/models/cesm1.0/pop2/doc/sci/POPRefManual.pdf.
Swingedouw, D., J. Mignot, P. Ortega, M. Khodri, M. Menegoz, C. Cassou, and V. Hanquiez, 2017: Impact of explosive volcanic eruptions on the main climate variability modes. Global Planet. Change, 150, 24–45, https://doi.org/10.1016/j.gloplacha.2017.01.006.
Tilmes, S., and Coauthors, 2018: CESM1(WACCM) Stratospheric Aerosol Geoengineering Large Ensemble Project. Bull. Amer. Meteor. Soc., 99, 2361–2371, https://doi.org/10.1175/BAMS-D-17-0267.1.
Weijer, W., W. Cheng, O. A. Garuba, A. Hu, and B. T. Nadiga, 2020: CMIP6 models predict significant 21st century decline of the Atlantic meridional overturning circulation. Geophys. Res. Lett., 47, e2019GL086075, https://doi.org/10.1029/2019GL086075.
Xie, M., J. C. Moore, L. Zhao, M. Wolovick, and H. Muri, 2022: Impacts of three types of solar geoengineering on the Atlantic meridional overturning circulation. Atmos. Chem. Phys., 22, 4581–4597, https://doi.org/10.5194/acp-22-4581-2022.
Zhang, Y., and Coauthors, 2023: Introducing a comprehensive set of stratospheric aerosol injection strategies. Egusphere, 2023, 1–32, https://doi.org/10.5194/egusphere-2023-117, in press.
Zhu, J., and Coauthors, 2021: Assessment of equilibrium climate sensitivity of the Community Earth System Model version 2 through simulation of the Last Glacial Maximum. Geophys. Res. Lett., 48, e2020GL091220, https://doi.org/10.1029/2020GL091220.
