1. Introduction
Global warming induced by an increase in atmospheric greenhouse gases is known to accompany a pattern of surface warming contrast between land and ocean regions: surface air temperature (SAT) increases more over land than over ocean. This land–ocean warming contrast is a striking feature of global warming, identified in observations over the recent past and in future projections using general circulation models (GCMs; Lambert and Chiang 2007; Manabe et al. 1991; Sutton et al. 2007, hereafter S07). Future changes in the land–ocean warming contrast could influence the regional climate via changes in atmospheric circulation (Kamae et al. 2014).
While the land–ocean warming contrast is a robust feature of global warming in projections, its physical mechanism has not yet been fully clarified. Intuitively, the land–ocean warming contrast can be explained by the large heat capacity of the ocean, which warms much slower than land. However, previous studies have shown that differences in heat capacity between land and ocean, represented by ocean heat uptake, are not the sole cause of the land–ocean warming contrast because the land surface warms more than the ocean surface even in an equilibrium state estimated by the CO2 doubling experiment using global atmosphere models coupled to a slab mixed-layer ocean (Manabe et al. 1991; S07). Several physical processes that contribute to land–ocean warming contrast have been suggested, but a comprehensive understanding has not yet been obtained (S07; Joshi et al. 2008; Byrne and O’Gorman 2013; Dong et al. 2009).
Similarly, a careful examination of the land–ocean warming contrast between the equilibrium response and transient climate response (TCR) is required. S07 showed that the land SAT increase was 50% ± 13% larger than the ocean SAT increase in the multimodel simulations with a CO2 increase of 1% per year, the experiment referred to as 1pctCO2, and that the ratio changed little over time. Their results led to the assumption that the ratio between land SAT change and ocean SAT change, ϕ, is constant regardless of the time evolution of the CO2 increase (Lambert et al. 2011; S07). Lambert et al. (2011) suggest that atmospheric energy transport between land and ocean plays a crucial role in determining ϕ. However, it made an assumption that effective radiative forcing (ERF) is spatially uniform, and furthermore, physical processes that alter the atmospheric energy transport remain unclear. Therefore, we focused on the time evolution of ϕ in transient climate change simulations and attempted to achieve a physically based understanding of the land–ocean warming contrast including climate adjustment processes and physical processes driving the energy transport anomaly.
In Part I of this study (Toda et al. 2021, hereafter Part I), we developed an energy budget framework to diagnose ϕ, based on two energy budget equations: one at the top of atmosphere (TOA), and another for the atmospheric column, the latter defined as a difference in energy between TOA and the surface (see section 2c). We then applied the method to the abrupt CO2 quadrupling experiment (hereafter referred to as abrupt4×CO2) using 15 Coupled Model Intercomparison Project phase 6 (CMIP6) GCMs. In the near-equilibrium state (years 121–150, hereafter NEQ), the land–ocean warming contrast is mainly caused by the difference in heat capacity and ERF between land and ocean, which seems to be explained by cloud reduction over land in the climate adjustment process. The same adjustment process can be a dominant cause for the land–ocean warming contrast in the equilibrium state (climate state with a global mean TOA radiative flux of 0 obtained by extrapolation for the global mean TOA, hereafter EQ). In Part II of this study, i.e., the present work, we apply the energy budget framework to the CMIP6 1pctCO2 experiments, which is a more realistic scenario with gradual increase in CO2 concentrations than abrupt4×CO2, and compare the mechanisms of the land–ocean warming contrast in TCR and NEQ, the latter examined in Part I. In addition, temporal variations in land–ocean warming contrast in 1pctCO2 and abrupt4×CO2 are investigated.
In section 2, we describe the models and numerical experiments, an index that measures the land–ocean warming contrast, and the energy budget framework. The physical mechanisms of the land–ocean warming contrast in TCR are analyzed and compared to those in NEQ in section 3. In section 4, the time evolution of the land–ocean warming contrast and associated changes in the surface heat flux in both 1pctCO2 and abrupt4×CO2 were investigated. In section 5, we demonstrate that compensation between processes that equally contribute to the land–ocean warming contrast explains the small temporal change in both experiments. Section 6 provides a summary and a discussion of the results.
2. Methods
a. Models, experiments, and index of the land–ocean warming contrast
In Part I, we focused on the land–ocean warming contrast at the near-equilibrium state. On the other hand, in Part II, we focus on that in TCR and the temporal change of ϕ in response to various CO2 increase scenario by analyzing the 1pctCO2 simulations and comparing them with the abrupt abrupt4×CO2 experiments. We analyze a single member (r1i1p1f1) of both experiments using 15 CMIP6 models, which were available when we started the analysis (Table 1). The experimental settings of both runs are the same except for the time evolution of the CO2 concentration.
Correlation coefficient between ΔN and ΔT in each of the CMIP6 abrupt4×CO2 experiments. The value statistically significant at the 95% level is shown by asterisks, indicating that all values are statistically significant.
The climate response, denoted as Δ, is defined by deviations in abrupt4×CO2 and 1pctCO2 from the corresponding preindustrial control experiment (piControl). We use the annual mean fields, and all data are regridded to 2.5° × 2.5° by linear interpolation before the analysis.
The land–ocean warming contrast, ϕ, is evaluated using an index used in Part I, defined as ϕ = ΔTL/ΔTO, where ΔTL and ΔTO are land-mean and ocean-mean SAT changes, respectively. In this article, the subscripts L, O, and G represent the land, ocean, and global mean, respectively. Antarctica is included when calculating the land mean, and the sea ice area is included in the ocean mean.
b. The energy budget framework
3. The land–ocean warming contrast in TCR
In many studies, TCR is defined as a 20-yr mean centered at year 70, when the atmospheric CO2 concentration is doubled from the preindustrial level (IPCC 2021). In this study, we redefine TCR as a 20-yr mean centered at year 140, when the CO2 level is quadrupled so that ERF matches abrupt4×CO2.
A scatterplot of ϕ between TCR and NEQ from the 15 CMIP6 models is presented in Fig. 1a, which shows that ϕ is highly correlated (r = 0.98) between them. This indicates that models with a large ϕ in NEQ tend to have a large ϕ in the TCR. A similar but slightly weaker relationship is found between the TCR and EQ (Fig. 1b, r = 0.78). However, ϕ in TCR tends to be systematically larger than ϕ in NEQ and EQ, as shown in Figs. 1a and 1b. In the multimodel mean, ϕ in TCR is 4% and 15% larger than ϕ in NEQ and EQ, respectively, although the difference between TCR and NEQ is not statistically significant in 95% confidence level. This result suggests that TCR is a climate state farther from the equilibrium state than NEQ. Indeed, the global-mean change in the TOA radiation, which measures the distance from equilibrium, is zero in EQ by definition, but large in TCR compared with NEQ (Fig. 1c).
(a) Scatterplot of ϕ in TCR against ϕ in NEQ. (b) As in (a), scatterplot of ϕ in TCR against ϕ in EQ. Each symbol indicates the model listed in Table 1. (c) The multimodel mean of ΔNG in TCR, NEQ, and EQ. Error bars indicate one standard deviation.
Citation: Journal of Climate 36, 13; 10.1175/JCLI-D-22-0483.1
The reconstructed land–ocean warming contrast, ϕe, for TCR is compared with that for NEQ and EQ (Fig. 2, the values of ϕe for TCR are presented in Table 2, and those for NEQ and EQ are in Tables S2 and S3. The values used for calculating ϕe in TCR are given in Table S1, and those for NEQ and EQ are in the supplementary material in Part I). As in NEQ and EQ, ϕe reproduces ϕ in the TCR very well, indicating that the framework works. By definition, the contribution of climate feedback (ϕλ) in TCR is the same as that of NEQ and EQ. Likewise, ERF is common among the three states, but the relative contribution to ϕ is the largest in TCR and the smallest in EQ. The major differences among TCR, NEQ, and EQ are observed in the contribution of energy transport, which is large negative in TCR and small positive in EQ, and the contribution of the land/ocean heat capacity difference, which is largely positive in TCR and zero in EQ. The largest positive contribution of the heat capacity in TCR is easy to understand because TCR is far from equilibrium and ocean heat uptake continuously takes place. In the abrupt4×CO2 run, the energy transport from land to ocean acts to weaken ϕ in the NEQ, but the direction of the transport changes in the EQ. Land-to-ocean heat transport is even larger in the TCR. Overall, the contributions of these two effects (horizontal energy transport and ocean heat uptake) tend to compensate for each other, leading to a similar value of ϕe among the three states. Section 5 examines the compensation mechanism.
Attribution of ϕ for TCR (orange bars), NEQ (blue bars), and EQ (sky blue bars). The left two bar groups show ϕ calculated using the simulated SAT and the reconstruction using Eq. (1) (ϕe), respectively. The other bars are the contribution of each term to reconstructed ϕ: base (equal to one, ϕ0), climate feedback (ϕλ), ERF (ϕF), atmospheric energy transport anomaly (ϕE), ocean heat uptake (ϕU), and the covariance term (ϕcov). Shaded bars indicate the multimodel mean, and error bars show the one standard deviation.
Citation: Journal of Climate 36, 13; 10.1175/JCLI-D-22-0483.1
The ϕ in individual CMIP6 models for TCR and its decomposition based on the energy budget framework.
In Fig. 2, intermodel spreads for each term are large, suggesting model uncertainty. However, the intermodel spreads of the three major terms (ERF, climate feedback, and the energy transport anomaly) tend to cancel each other, causing the spread of ϕ to be small and the multimodel mean value to be robust (cf. section 4d of Part I). In NEQ, intermodel spread of ϕ is well explained by that of λL (cf. section 4e of Part I). The high correlation between ϕ in NEQ and TCR suggests (Fig. 1a) that the intermodel spreads of ϕ in TCR are explained by the same manner as NEQ.
4. Time evolution of land–ocean warming contrast in the 1pctCO2 run
a. Climate response
Prior to investigating the time evolution of ϕ in the 1pctCO2 run, the basic features of the temporal evolution of the climate response in the 1pctCO2 run are described. The time series of ΔTL and ΔTO together with several variables in the energy budget equations are shown in Figs. 3a–d. Unlike the abrupt4×CO2 run (Fig. 3e), the SAT response is very small in the initial years when the ERF is small, hindering the robust estimation of the land–ocean warming contrast during the initial decades, but eventually become clear as the ERF increased (Fig. 3a). The time series of the TOA net radiation imbalance, ΔNL and ΔNO, shows that the climate system gains excess energy both over land and ocean, but ΔNO is slightly larger than ΔNL after CO2 doubling in year 70 (Fig. 3b). As in the abrupt4×CO2, the response of the net surface energy flux over land, ΔUL, is nearly zero after the initial years because of the small heat capacity (Fig. 3c). In contrast, the net surface energy flux over the ocean, ΔUO, continuously increases (negative sign denoting downward flux), indicating that ocean heat uptake operates throughout the period of the 1pctCO2 run. The sign of the energy transport is opposite between land and ocean, which indicates that excess energy is always transported from land to ocean (Fig. 3d). In the 1pctCO2 run, ocean heat uptake as represented by ΔUO and the energy transport from land to ocean intensified with time, unlike the abrupt4×CO2 run, which showed the weakening of these fluxes.
(a)–(d) Time series of changes in (a) SAT ΔT, (b) TOA radiation ΔN, (c) surface net energy flux ΔU, and (d) atmospheric energy transport ΔK, over land (red) and ocean (blue) in the CMIP6 1pctCO2 experiments. Thick curves indicate the multimodel mean and the shading denotes the 90% range. (e)–(h) As is (a)–(d), but for the abrupt4×CO2 run [(f)–(h) are reprinted from Fig. 6 in Part I].
Citation: Journal of Climate 36, 13; 10.1175/JCLI-D-22-0483.1
The climate response in the abrupt4×CO2 run can be separated into two time scales (Knutti and Hegerl 2008): a fast response that occurs within a few months after the CO2 concentration quadruples, and a slow response that occurs over a longer time scale of decades while the ocean surface warms. The fast response is regarded as climate change associated with rapid adjustment processes responsible for generating ERF, whereas the slow response consists of climate feedback and ocean heat uptake. The response at the two time scales can be separated using regression analysis with respect to ΔTO because the fast response does not change over time owing to a fixed CO2 concentration. The fast and slow climate responses decomposed from the abrupt4×CO2 run are schematically shown in Fig. 4 (reprinted in Fig. 7, Part I). In the fast response, ΔTO is zero, but ΔTL can increase slightly (1.21 ± 0.28 K) because of the small heat capacity of the continents. As a result, ΔNL in the fast response (6.76 ± 0.72 W m−2) is smaller than ΔFL (7.84 ± 0.99 W m−2, see Table S1) owing to the feedback process induced by ΔTL. Signs of all fluxes were opposite for the fast and slow responses (Fig. 4); therefore, ocean heat uptake and energy transport from land to ocean weakens over time in the abrupt4×CO2 run because only the slow response changes over time (Figs. 3g,h).
Summary of energy budget changes in (a) the fast response to quadrupled CO2 before ocean surface temperature (ΔTO) starts to increase and (b) slow response corresponding to climate changes when ΔTO increases by 1 K. In (a), all units of the values attached to the arrows are W m−2, and units of ΔTL and ΔTO are K. In (b), all units of the values attached to the arrows are W m−2 K−1, and units of ΔTL and ΔTO are K K−1 (reprinted from Part I Fig. 7).
Citation: Journal of Climate 36, 13; 10.1175/JCLI-D-22-0483.1
During the 1pctCO2 run, the climate change signal at a time is likely a combination of fast and slow responses, which are explicitly inseparable. However, we demonstrated that these responses in 1pctCO2 can be decomposed using the results from abrupt4×CO2, following the method described below.
(a)–(e) Time series of raw values (black) and estimated values (brown) of ΔTL, ΔNL, ΔNO, ΔUL, and ΔUO based on Eq. (12) in 1pctCO2. Thick curves indicate the multimodel mean and the shading denotes the 90% range.
Citation: Journal of Climate 36, 13; 10.1175/JCLI-D-22-0483.1
Equation (12) can also be used to reproduce the horizontal distribution of climate responses in the 1pctCO2 run by replacing
(a),(c),(e) Horizontal map of ΔT, ΔN, ΔU in TCR, obtained from the CMIP6 multimodel mean response. (b),(d),(f) Reproduced horizontal map of ΔT, ΔN, ΔU in TCR estimated from Eq. (12), obtained from the CMIP6 multimodel mean response. Dots in (c)–(f) indicate locations where the sign is significant in a t test at the 95% significance level. Signs in (a) and (b) are statically significant in all regions.
Citation: Journal of Climate 36, 13; 10.1175/JCLI-D-22-0483.1
Given that the climate response in the 1pctCO2 run can be expressed as a linear combination of fast and slow responses in the abrupt4×CO2 run, increasing trends of ocean heat uptake and energy transport (Figs. 3c,d) are interpreted as a consequence of the fast response overwhelming the slow one. Despite an apparent difference in the time evolution of SAT changes between the 1pctCO2 and abrupt4×CO2 runs (Figs. 3a–h), the physical processes responsible for these responses can be understood in a unified manner. The strength of the fast response is the same for TCR and NEQ, but the strength of the slow response is weaker in TCR than in NEQ owing to a smaller ΔTO.
b. Comparison of time series of ϕ between 1pctCO2 and abrupt4×CO2
We now compare the time evolutions of the land–ocean warming contrast between 1pctCO2 and abrupt4×CO2. The time series of ϕ in both experiments is shown in Figs. 7a and 7b (annual-mean values in Fig. 7a and their low-frequency component by taking a 10-yr moving average in Fig. 7b). The overall evolution is similar between the two experiments: an initial large value of ϕ followed by a decrease toward a constant value of approximately 1.5. However, the model spread is quite large in 1pctCO2 until about 70 years because ΔTO is small, and therefore, ϕ is perturbed by interannual fluctuations (Fig. 7a). As global warming evolves, the model spread is reduced, and the multimodel mean value of ϕ approaches that in the abrupt4×CO2 run.
(a) Time series of multimodel mean of ϕ in the abrupt4×CO2 run (black line) and the 1pctCO2 run (green line). (b) As in (a), but time series are 10-yr moving averaged. The color shading is the 90% interval. (c) Ten-year moving averaged time series of multimodel mean of ϕ (black) and ϕe (purple) in the abrupt4×CO2 run. (d) As in (c), but for the 1pctCO2 run. The purple line showing the multimodel mean for ϕe has thick and thin sections. The thick line section indicates the multimodel mean values of ϕ and ϕe are not statistically indistinguishable at that point in time with 95% significance level based on Student’s t test. Oppositely, the thin line section indicates the multimodel mean values of ϕ and ϕe are statistically significantly different at that point.
Citation: Journal of Climate 36, 13; 10.1175/JCLI-D-22-0483.1
Two common features are identified in the time series of ϕ. One is a rapid decay during the first 20 years, which then changes very slowly at a rate of 0.1 per century (Fig. 7b). This feature is found not only in the multimodel mean, but also in each model (Figs. S3a,b). The other is a similarity of the value during the late periods: ϕ = 1.49 ± 0.11 in abrupt4×CO2 and ϕ = 1.55 ± 0.11 in 1pctCO2. As mentioned in section 3, ϕ in TCR is only 4% larger than in NEQ. The reasons for the above similarity can be found from analyses using the energy budget framework.
The energy budget framework based on Eq. (7) is applied to the 1pctCO2 run to examine whether the nearly constant magnitude of the land–ocean warming contrast is reproduced in transient global warming. The 10-yr moving average of multimodel mean time series of ϕ and ϕe in 1pctCO2 and abrupt4×CO2 are presented in Figs. 7c and 7d. Hence, ϕe reproduces ϕ both in terms of the multimodel mean and the intermodel spread for the abrupt4×CO2 run (Fig. 7c, see also Figs. S3a,c for the individual models). However, the model spread of ϕe is too large in the 1pctCO2 run during the first half of the period when the denominator of Eq. (7) is extremely small, and the multimodel mean values of ϕ and ϕe are significantly different. However, after around year 70 when the global warming signal is sufficiently large, the model spread of ϕe becomes small and there is no significant difference between the multimodel mean values of ϕ and ϕe.
The small rate of temporal change during the last few decades is well reproduced in 13 out of 15 models (Figs. S3b,d). Two models, CESM and CESM-WACCM, show that ΔNO and ΔTO are not highly correlated, although the correlation is significant at the 95% confidence level (Table 1). Therefore, the energy budget framework works poorly. In the next subsection, the reason for the small temporal change in ϕ is further investigated.
c. Decomposed time evolution of the land–ocean warming contrast
The low-frequency time series of ϕe in the multimodel mean shown in Figs. 7c and 7d are attributed to individual terms (Fig. 8). The decomposition of NEQ is explained in Part I, but here the decomposed time series are shown for abrupt4×CO2 (Fig. 8a). As in the abrupt4×CO2 run, the positive contribution of the ERF to ϕe plays a major role in the 1pctCO2 run (blue curve in Fig. 8b). In addition, the contribution of heat uptake (pink curve) is an important factor for the land–ocean warming contrast in the 1pctCO2. Atmospheric energy transport (cyan curve) acts to weaken the land–ocean warming contrast by transporting energy from land to ocean over 150 years in both experiments. All terms decay over time and change little after approximately 70 years. The rate of decay for the energy transport contribution is somewhat larger in abrupt4×CO2 than in 1pctCO2, resulting in a smaller negative contribution to NEQ (Fig. 2). Although there is a slight difference in the time evolution of each contribution between the two experiments, the land–ocean warming contrast is commonly explained by the positive contribution of heat capacity (associated with ocean heat uptake) and ERF differences, partly compensated by the negative contribution of horizontal energy transport.
(a) Multimodel mean of time series of ϕ and each contribution for the abrupt4×CO2 run. (b) As in (a), but for 1pctCO2. The ϕEU is the combined contribution of energy transport and heat capacity defined in Eq. (17).
Citation: Journal of Climate 36, 13; 10.1175/JCLI-D-22-0483.1
Although the equations are not shown, a similar argument holds for the decay of the energy-transport contribution to ϕ. Despite an increase in the energy transported by the atmosphere, it does not exceed the increasing amount of radiative energy used for increasing SAT over land; therefore, the contribution to ϕ weakens over time. The sluggish growth in energy transport could probably be explained by the ocean heat uptake weakening. The physical link between them is examined in detail in section 5.
A question arises as to whether there is a physical link between the contributions of energy transport and ocean heat uptake. As discussed in section 4b, the climate response in 1pctCO2 can be decomposed into fast and slow responses obtained from the results of the abrupt4×CO2 run (Fig. 4). Therefore, the possibility of a physical link between energy transport and ocean heat uptake should be examined at the respective time scale.
5. Processes linking ocean heat uptake and energy transport
We have shown that the compensation between the atmospheric energy transport and the ocean heat uptake is important in suppressing the temporal changes in ϕ, consistent with Lambert et al. (2011). In this section, we newly discuss physical mechanisms that link these two processes.
During the fast response without warming over the ocean surface, the ERF is the driver of rapid SAT change over land. The SAT increase is rapid, but much of the excess energy is transported to the ocean, which is then absorbed beneath the ocean surface (Fig. 4), corresponding to the ocean heat uptake. Energy transport from land to ocean acts to suppress land warming and therefore indicates a negative contribution to ϕ. In contrast, ocean heat uptake slows ocean surface warming and positively contributes to ϕ. In the slow response, in which the ocean surface eventually warms, excess energy associated with ΔTO is transferred to the atmosphere from the surface. This energy flux corresponds to a weakening of the ocean heat uptake. Approximately 30% of the input energy from the ocean surface is transported to land, and as a result, ΔNL and ΔNO show similar values, indicating energy loss over land and ocean at a similar rate (Fig. 4b).
Thus, atmospheric energy transport seems to link energy transport and ocean heat uptake at both fast and slow time scales. We first focus on the slow response (section 5a), followed by a discussion on the fast response (section 5b) obtained from abrupt4×CO2 (see section 4a). The time evolution of the climate response in 1pctCO2 is a combination of these, and is discussed in section 5c.
a. Slow response
Figure 9a shows the horizontal distribution of SAT response regressed on ΔTO. This represents the pattern of the slow response and resembles the global warming pattern in future projections (IPCC 2021): polar warming amplification, El Niño–like warming in the tropical Pacific, and land–ocean warming contrast. A corresponding horizontal map of the net surface flux response (ΔU) is shown in Fig. 9b. It is easily found that the upward surface flux occurs where the ocean surface warming is larger than the surrounding region, corresponding to the upward surface energy flux anomaly of 1.4 ± 0.43 W m−2 with increase in ΔTO by 1 K (Fig. 4b). This upward ocean surface energy flux anomaly corresponds to the weakening of ocean heat uptake, where the ocean temperature increases. The horizontal distribution of the TOA net energy imbalance (ΔN) is relatively uniform, unlike that of the surface flux response (Fig. 9c). There is clear spatial coherence between ΔU and the divergence of the atmospheric energy transport response (∇ ⋅ ΔE) (Fig. 9d). In particular, both were large over the tropical Pacific Ocean. The vectors in Fig. 9d represent the divergent component of the atmospheric energy flux (ΔE), showing that the energy exchange between land and ocean occurs mainly in the tropics, unlike its climatology, which shows divergent fluxes from low to high latitudes (see Fig. 3a in Part I).
Horizontal map of (a) SAT anomaly ΔT, (b) surface flux anomaly ΔU, (c) TOA flux anomaly ΔN, and (d) divergence of energy transport anomaly ∇ ⋅ ΔE when ΔTO increase by 1 K (slow response), obtained from the CMIP6 multimodel mean response. Values are calculated by the regression coefficient of the linear regression of ΔTO in each model and averaging them up. The vectors in (d) indicate divergence component of ΔE in the multimodel mean. Dots indicate locations where the sign is significant in a t test at the 95% significance level.
Citation: Journal of Climate 36, 13; 10.1175/JCLI-D-22-0483.1
The fractional contributions of the energy flux divergence over land and ocean, ΔKL and ΔKO, at three latitude bands to their global mean values are summarized in Figs. 10a and 10b. In the multimodel mean (bars), 57% of the net land energy flux convergence occurs in the tropics (Fig. 10a), which is 1.3 times larger than the area fraction of the tropics (44%), indicating that energy is intensively transported to tropical land regions. The fractional contribution of the northern extratropics (30°–90°N) is only 27%, which is much smaller than the area occupied (43%), indicating weak energy transport to land over mid–high latitudes in the Northern Hemisphere (Fig. 10a).
(a) ΔKL in the slow response (gray bar), and the contribution of each latitude band (red bar). (b) As in (a), but for ΔKO. The contribution of each latitude band is represented in blue bar. Fractional area coverage at each latitudinal band is represented in orange bar. (c) The horizontal distribution of ∇ ⋅ΔE over ocean (color) and the relative contribution of each region to ΔKO in the multimodel mean. (d) As in (c), but for ΔKL.
Citation: Journal of Climate 36, 13; 10.1175/JCLI-D-22-0483.1
During the slow response, 78% of ocean energy flux divergence occurs in the tropics, in contrast to the very small divergence over the northern extratropical oceans (Fig. 10b). Thus, energy transport to land over the Northern Hemisphere high latitudes is achieved from the tropical oceans. Figure 10c shows the fractional contribution rate of the net ocean energy flux divergence (ΔKO) to the global mean value in each ocean basin. A considerable amount of ocean energy flux divergence (112%) occurred in the tropical Pacific, which is nearly twice as large as the contribution from all other regions. The Atlantic high latitudes absorb energy, indicating ocean heat uptake associated with the overturning circulation in the slow response. For land energy flux convergence, there is no prominent convergence (Fig. 10d).
The energy exchange in the tropics, which is the main factor of energy flux divergence and convergence, is further investigated (see supplementary chapter B). The energy exchange is explained by the weakening of the Walker circulation driven by enhanced convective activities over the tropical Pacific Ocean associated with upward ocean surface flux (Figs. S6, S7a, and Figs. 11a,b) and these mechanisms are common in all 15 models. Indeed, the horizontal distribution of ∇ ⋅ ΔE and that of ω500 are in good agreement in the tropics (Figs. 9d and 11b). The slowdown of tropical circulation is a robust signal of global warming, as detected in previous studies (Held and Soden 2006; Vecchi and Soden 2007). In other words, the major part of the land–ocean energy exchange in the slow response is driven by the robust response of tropical circulation in the atmosphere.
The horizontal distribution of the vertical component of velocity in pressure coordinates (ω) at the 500 hPa in (a) climatology of piControl, (b) the slow response, and (c) the fast response. Dots indicate locations where the sign is significant at the 95% significance level with the t test.
Citation: Journal of Climate 36, 13; 10.1175/JCLI-D-22-0483.1
Although the main energy exchange occurs within the tropics, some energy is transported poleward from the tropics. Energy from the tropical Pacific is likely to be transported to mid- and high latitudes by eddy activity. However, the investigation of energy transport by eddies requires daily data, and it is technically impossible to obtain daily values for the slow response because a slow response is calculated by linear regression analysis based on annual mean values. Future research is required to understand the physical processes by which energy dissipated from the tropical Pacific Ocean is transported to land at mid- and high latitudes.
b. Fast response
Similar to the procedure for estimating the slow response, the horizontal patterns of the fast response are obtained using linear regression with ΔTO (Fig. 12, showing the y intercepts, not the regression coefficients). A comparison of the energy flux changes at the TOA and surface and atmospheric energy transport shows that the latter is mainly driven by the change in surface energy flux over the ocean. By definition, however, the surface flux change associated with the fast response is not related to changes in ocean surface temperature. A possible process that drives the rapid change in ocean surface fluxes is rapid tropospheric adjustment (Andrews et al. 2009). Namely, a weakening of the atmospheric longwave cooling directly induced by the increase in CO2 must be balanced by a reduction in the surface latent heat flux from the ocean surface. Indeed, the global-mean value of the fast response of ΔUO (−8.40 ± 0.80 W m−2) is dominated by the latent heat flux (−4.30 ± 0.58 W m−2), consistent with the above energy budget argument. The rest is explained by the longwave radiation (−3.81 ± 0.50 W m−2), shortwave radiation (1.00 ± 1.11 W m−2), and the sensible heat flux (−1.29 ± 0.46 W m−2). The downward longwave radiation corresponds to instantaneous CO2 radiative forcing.
Horizontal map of (a) surface flux anomaly ΔU, (b) divergence of energy transport anomaly ∇ ⋅ ΔE, and (c) TOA flux anomaly when ΔTO = 0 (the fast response), obtained from the CMIP6 multimodel mean response. Values are calculated by the y interceptions of the linear regression of ΔTO in each model and averaging them. The vectors in (b) indicate divergence component of ΔE in the multimodel mean. Dots indicate locations where the sign is significant in a t test at the 95% significance level.
Citation: Journal of Climate 36, 13; 10.1175/JCLI-D-22-0483.1
The horizontal patterns of the individual components of the fast response of ΔUO are presented in Fig. 13. The change in longwave radiation is highly uniform in space (Fig. 13a), and the pattern of ΔUO (Fig. 12a) is explained by the shortwave radiation (Fig. 13b) and latent heat flux (Fig. 13c). The horizontal distribution of shortwave radiation is in good agreement with that of cloud cover fraction (Fig. 13e). The quite large cloud fraction decrease found over tropical Pacific in Fig. 13e can be explained by the large horizontal moisture divergence (Fig. 13f, the calculation method of the moisture divergence is in supplementary chapter C). This moisture divergence corresponds to the subsident flow region in the fast response (Fig. 11c), because in subsident flow region, atmospheric circulation diverges at the lower atmospheric layers with high water vapor content. In the fast response, cloud cover decreases at moisture divergence zones since there is not an adequate supply of water vapor by evaporation over ocean.
The horizontal distribution of (a) surface long wave flux, (b) surface short wave flux, (c) latent heat flux, (d) sensible heat flux in the fast response, (e) cloud area fraction, and (f) the divergence of moisture flux calculated based on Eq. (S12) (color) plotted with the divergent component of moisture fluxes (vector). Dots indicate locations where the sign is significant in a t test at the 95% significance level.
Citation: Journal of Climate 36, 13; 10.1175/JCLI-D-22-0483.1
Figure 14a shows that 55% of the energy divergence over land occurred in the tropics, and this contribution rate was larger than the area fraction (44%). However, the mid- to high latitudes of the Northern Hemisphere show smaller energy divergence (34%, Fig. 14a) than its area fraction (43% of the total land). Thus, it can be concluded that tropical land is the main energy divergence zone. However, the tropical ocean mean energy convergence is small (Fig. 14b). This is because the tropical Atlantic and Indian Ocean, where ocean heat uptake is small, are energy divergence zones, whereas the tropical Pacific Ocean shows large energy convergence associated with large ocean heat uptake (Fig. 14c). Figure 14c shows that the main energy convergence zones are the tropical Pacific Ocean and the Southern Ocean (the combined contribution rate to net ocean convergence is 96% in the multimodel mean). This indicates that heat absorption in the deep-water formation region in the Southern Ocean is very important for the fast response of ocean heat uptake.
(a) ΔKL in the fast response (gray bar), and the contribution of each latitude zone (red bar). (b) As in (a), but for ΔKO. The contribution of each latitude band is represented in blue bar. Fractional area coverage at each latitudinal band is represented in orange bar. (c) The horizontal distribution of ∇ ⋅ΔE over ocean (color) and the relative contribution of each region to ΔKO in the multimodel mean. (d) As in (c), but for ΔKL.
Citation: Journal of Climate 36, 13; 10.1175/JCLI-D-22-0483.1
Convergence from tropical land to the tropical Pacific is significant, although its contribution is lower than the energy exchange within the tropics in the slow response. By applying the same analysis as the slow response, the energy transport from tropical land to the tropical Pacific Ocean is explained by the strengthening of the Walker circulation, as opposed to the slow response (see supplementary chapter B).
In the fast response, energy transport from tropical land to the Southern Ocean is another main energy exchange mechanism (see supplementary chapter B). This energy transport can be explained by the meridional overturning circulation and eddy activities such as baroclinic instabilities. In addition to the slow response, energy transport in the midlatitudes can be an important future issue. To analyze the fast response, the CO2-forced, fixed SST experiment is also useful. Lambert et al. (2011) examined the atmospheric energy transport in the doubled CO2, fixed SST experiment with single model, and its direction is from land to ocean, qualitatively consistent with our results.
c. Combined response
In the previous subsections, we have shown that ocean heat uptake and atmospheric energy transport are dynamically linked at both fast and slow time scales. Therefore, the cancelation between the contribution of heat capacity (ϕU) and energy transport (ϕE) and consequently a small temporal change in ϕ can be explained on a physical basis.
In the 1pctCO2 experiment, the surface energy flux ΔUO, increased downward with time (Fig. 3c), which implies that the ocean interior continuously absorbed heat. In addition, the atmospheric energy flux is divergent over land and convergent over the ocean throughout the period, and is enhanced over time. They include both fast and slow responses, as we have explained so far, and their relative role in time evolution is investigated here. Figure 15 shows the multimodel mean surface flux response in the TCR obtained from the multimodel mean
The reproduced multimodel mean horizontal distribution of surface flux in TCR by using surface flux distribution of multimodel mean the fast response and the slow response. Points where the contribution of the fast response to the sea surface flux is greater than 50% are hatched in purple. Points where the contribution of the slow response to the sea surface flux is greater than 50% are hatched in green. The points where surface flux is less than 2(W m−2) are not hatched.
Citation: Journal of Climate 36, 13; 10.1175/JCLI-D-22-0483.1
There is an interesting correspondence between the sign of ΔUO and the relative contributions of the fast and slow responses. In the southern tropics (0°–30°S), weakening of heat uptake associated with the slow response prevails; therefore, ΔUO is positive (upward). However, in many ocean regions in the extratropics, particularly the North Atlantic and Southern Oceans, excess heat is absorbed; that is, ΔUO is negative (downward) explained primarily by the fast response. The fast response overwhelms the slow response overall, except for the tropical ocean, where the slow response plays a role.
During the 1pctCO2 run, both the fast and slow responses vary with time each year, but the predominance of the fast response each year enhances ΔUO and land-to-ocean energy transport. In contrast, the predominance of the slow response over tropical oceans weakens the ocean heat absorption and becomes an energy divergence field. Thus, the annual mean contributions of the fast and slow responses cancel each other out in the tropics, so that the temporal changes in 1pctCO2 energy transport and ocean heat absorption are not explained by the Walker circulation mechanism within the tropics from an annual mean perspective, but rather the importance of energy transport from the tropics to the mid–high latitudes is suggested.
In contrast to the 1pctCO2 run, the slow response alone varies with time in the abrupt4×CO2 run. Therefore, energy transport and ocean heat uptake are linked through the weakening of Walker circulation. These results suggest that the physical mechanisms that link temporal change in energy transport and ocean heat uptake change depending on the combination of fast and slow responses, in other words, depending on the global warming scenario.
d. Summary of the mechanisms for the constant ϕ
Physical mechanisms of land–ocean warming contrast in a transient climate response are summarized as follows: As in the steplike CO2 quadrupling experiment (abrupt4×CO2), climate change signals in the 1pctCO2 run can be decomposed into a fast response directly associated with increasing CO2 concentration and a slow response associated with warming of the ocean surface. The difference from the abrupt4×CO2 run is that both the responses evolve over time. The changes in the energy fluxes at these time scales and their combinations are schematically shown in Fig. 16.
Schematic diagram of the multimodel mean the land–ocean warming contrast in transient state. In the 1pctCO2 run, (a) the fast response and (b) the slow response occurs every year. (c) The sum of them is schematically represented, indicating that the fast response overcomes the slow response. The enhanced ERF over land due to stomatal resistance and moisture transport from land contributes to the land–ocean warming contrast every year as well as near-equilibrium state (label 1). Land to ocean energy transport works to reduce ΔTL (label 2) and oppositely ocean heat uptake suppresses ΔTO (label 3). The contribution of ERF is not time varying (ϕF ≈ 0.4). The contribution of energy transport and heat uptake is time varying, but, since the two contributions cancel each other out and change in tandem, the combined contribution will have a smaller temporal change (ϕUE ≈ 0.2, label 4).
Citation: Journal of Climate 36, 13; 10.1175/JCLI-D-22-0483.1
The mechanism of the land–ocean warming contrast identified in the multimodel mean fields is described below. As the CO2 concentration increases, the ERF is always greater over land than over the ocean, which is part of the fast response. As the response occurs with the same mechanism as for the near-equilibrium state, this ERF difference is likely due to the decrease in low-level clouds (label 1 in Fig. 16c) probably due to stomatal resistance over land (Dong et al. 2009; Doutriaux-Boucher et al. 2009). However, some of the energy is transported from land to the ocean and acts to suppress the temperature increase over land (Fig. 16a and label 2, Fig. 16c). In the fast response, the response of the net surface energy flux is downward, which indicates that excess heat is taken up by the ocean interior, which acts to suppress the temperature increase over the ocean (label 3, Fig. 16c). As the ocean surface warms, both the land to ocean energy transport and ocean heat uptake weakens as a slow response (Fig. 16b).
Overall, the fast response overcomes the slow response throughout the experiment for 150 years, resulting in the net energy transport from land to ocean that acts to weaken the land–ocean warming contrast, and the ocean heat uptake that acts to strengthen it (Figs. 16c and 8b). The contribution of the ERF to the land–ocean warming contrast does not vary with time (ϕF ≈ 0.4) and is the largest factor for ϕ. Temporal changes in the contributions of energy transport and heat uptake are large, but they are physically linked and tend to cancel each other out via changes in the atmospheric circulation and eddy (label 4 in Fig. 16c). Thus, the combined contribution of the two is of secondary importance (ϕEU ≈ 0.2). As a result, ϕ in TCR is approximately 1.55 ± 0.11, which is approximately 4% larger than that in NEQ, and the value does not change much after the global warming signal becomes large enough during the latter half of the 1pctCO2 run.
6. Summary and discussion
In Part II of this study, we investigated the mechanisms and characteristics of the land–ocean warming contrast (ϕ) in transient climate simulations by analyzing the results of the 1pctCO2 experiment using 15 CMIP6 GCMs. We used the energy budget framework proposed in Part I to evaluate the contribution of land–ocean differences in the ERF, climate feedback, heat uptake, and atmospheric energy transport to ϕ.
To compare with NEQ, we estimated ϕ in TCR by taking the 21-yr mean from 130 to 150 years when the CO2 concentration reached the same level as the abrupt4×CO2 run. In all models, ϕ was slightly larger in TCR than in NEQ by approximately 4%. This is due to a larger positive contribution of ocean heat uptake, indicating that the TCR is far from the equilibrium state, but the difference is not significant.
The time evolutions of the SAT response and energy fluxes show clear differences between the abrupt4×CO2 and 1pctCO2 runs. Nevertheless, the time series of climate response, including ϕ, in 1pctCO2 can be reproduced by referring to abrupt4×CO2, which is decomposed into a fast response associated directly with the increase in the CO2 concentration and a slow response associated with ocean surface warming. This indicates that the land–ocean warming contrast in the TCR can be understood using the mechanisms identified in Part I.
For the first 70 years in the 1pctCO2 runs, ϕ cannot be robustly estimated because the small global warming signal was perturbed by large interannual variability. After the signal became large enough against internal climate variability, the time series of ϕ in the 1pctCO2 run was stable and did not change over time (ϕ = 1.55 ± 0.11) similar to that in the abrupt4×CO2 run (ϕ = 1.49 ± 0.11). The near-constant ϕ in both experiments was primarily due to a constant (positive) contribution of the ERF. Furthermore, the contributions of atmospheric energy transport and ocean heat uptake act to compensate for each other, preventing ϕ from changing in time during the latter half of the 1pctCO2 experiment.
The physical linkage between atmospheric energy transport and ocean heat uptake, which explains the cancelation of ϕ, was investigated by means of fast and slow responses. At both time scales, energy transport occurred due to changes in atmospheric circulation excited by the change in ocean surface heat flux, and not by the change in the moist static energy between land and ocean. The surface heat flux changes over the ocean were strongly influenced by the ocean heat uptake; therefore, the contributions of the two processes to ϕ are physically linked through changes in large-scale atmospheric circulation.
Our results suggest that the temporal change in ϕ is small in any CO2 emission scenario as the cancellation of the contribution of ocean heat uptake and energy transport occurs in those experiments. However, in observations and realistic climate change simulations, radiative forcing factors other than CO2 (such as aerosols) vary over time. The extent to which the mechanisms identified in this study can be applied to the land–ocean warming contrast observed in the past and in future climate projections is not yet clear, and the question remains for future works.
In section 3, ϕ in EQ was significantly smaller than those in NEQ and TCR (see Fig. 2), consistent with previous studies such as S07 and Joshi et al. (2008). Intuitively, the small ϕ in the equilibrium state can be explained by the absence of the heat capacity term (ϕU). However, our analyses indicate that the process is not that simple; ϕU slowly decays toward equilibrium, but it is partly canceled by ϕE. The land–ocean ERF difference (ϕF) also decays in time, contributing to the small ϕ in EQ.
This study focused on land- and ocean-mean SAT responses to increasing CO2 concentration, and did not discuss regional changes among continents or ocean basins. However, our results provide some insights into the mechanisms of regional temperature changes, particularly over land. A useful implication obtained from Part II is that the horizontal distribution of SAT changes in transient warming simulations can be reproduced by summing the fast and slow responses obtained from the abrupt4×CO2 experiment. The distribution of SAT changes over land differed between the fast and slow responses (Figs. S2a,b), which affects regional temperature changes differently.
On one hand, the fast response shows that the temperature increase is weak over tropical land areas, likely due to a rapid energy exchange between land and ocean to minimize the ERF difference between them (Fig. S2a). In contrast, the middle and high latitudes of the Northern Hemisphere show very large temperature increases over land. This warming is induced by ERF, but excess energy is not transported to ocean areas. In the slow response, the distribution of SAT change was close to spatially uniform, except in the polar region (Fig. S2b). There are slight warming over tropical land areas as well, likely due to the weakening of energy transport from land to ocean as SAT increases over oceans.
The above arguments indicate that the SAT change mechanism, including the role of fast and slow climate responses, is different between the tropics and Northern Hemisphere extratropics. Over tropical land, ERF heating is not vital, but that by horizontal transport from the surrounding oceans is important for SAT increase. This is consistent with the weak temperature gradient in the tropical atmosphere. Over the Northern Hemisphere extratropics, the rapid ERF warming is important for the SAT increase over land, which is partly compensated by energy transport to the ocean. We investigated the fractional contribution of the fast response to the multimodel mean SAT change in TCR (Fig. S9). This result supports the above arguments: the fast response contributes to TCR by less than 20% over most tropical land areas, but contributes to TCR by more than 30% over the higher latitudes of the Northern Hemisphere.
Thus, the separation of surface warming signals in transient climate simulations into fast and slow components provides useful insight into the processes responsible for regional climate change. A similar separation may be possible for quantities other than SAT, and future studies will show the attribution and projections of regional climate changes from this perspective.
In the end, we discuss the relevance of this study to previous studies. Since the discussion of previous studies related to the land–ocean warming contrast at the equilibrium state was given in Part I (e.g., S07; Joshi et al. 2008), here we discuss the study on the contrast in TCR. Sejas et al. (2014, referred to as S14) has quantified the individual contributions of the CO2 forcing and other climate components to the land–ocean warming contrast in transient climate by using the climate feedback-response analysis method (CFRAM; Lu and Cai 2009; Cai and Lu 2009). They defined the doubled CO2 time point in the 1pctCO2 experiment of the NCAR CCSM4 as TCR. S14 has highlighted the importance of the ocean heat transport, ocean heat storage, and effects of clouds on shortwave radiation to ϕ > 1 in TCR in addition to the difference in latent and sensible heat flux between land and ocean whose importance is suggested in S07. Because the effect of clouds on shortwave radiation in CFRAM contains the land–ocean difference in cloud adjustment processes which is the main source of the contribution of ERF (ϕe), the conclusion in S14 is consistent with the present study, despite the use of different methods, which suggests that ocean heat uptake and difference in cloud adjustment process are important for the land–ocean warming contrast in TCR.
Acknowledgments.
We acknowledge the modeling groups, the PCMDI, and the WCRP’s WGCM for their efforts in making the CMIP6 multimodel dataset available. We also thank Kaoru Sato, Yu Kosaka, and three anonymous reviewers for providing valuable comments on the manuscript. M.T. was supported by a Grant-in-Aid from the Japan Society for the Promotion of Science (JSPS) Fellows (19J20697). M.W. and M.Y. were supported by the Integrated Research Program for Advancing Climate Models (JPMXD0717935457) and the Program for Advanced Studies of Climate Change Projection (SENTAN) (JPMXD0722680395) from the Ministry of Education, Culture, Sports, Science and Technology (MEXT), Japan.
Data availability statement.
The data used in this study are publicly available. The CMIP6 simulation data are available from the Earth System Grid Federation (https://esgf-node.llnl.gov/).
REFERENCES
Andrews, T., P. M. Forster, and J. M. Gregory, 2009: A surface energy perspective on climate change. J. Climate, 22, 2557–2570, https://doi.org/10.1175/2008JCLI2759.1.
Byrne, M. P., and P. A. O’Gorman, 2013: Land–ocean warming contrast over a wide range of climates: Convective quasi-equilibrium theory and idealized simulations. J. Climate, 26, 4000–4016, https://doi.org/10.1175/JCLI-D-12-00262.1.
Cai, M., and J. Lu, 2009: A new framework for isolating individual feedback processes in coupled general circulation climate models. Part II: Method demonstrations and comparisons. Climate Dyn., 32, 887–900, https://doi.org/10.1007/s00382-008-0424-4.
Dong, B., J. M. Gregory, and R. T. Sutton, 2009: Understanding land–sea warming contrast in response to increasing greenhouse gases. Part I: Transient adjustment. J. Climate, 22, 3079–3097, https://doi.org/10.1175/2009JCLI2652.1.
Doutriaux-Boucher, M., M. J. Webb, J. M. Gregory, and O. Boucher, 2009: Carbon dioxide induced stomatal closure increases radiative forcing via a rapid reduction in low cloud. Geophys. Res. Lett., 36, L02703, https://doi.org/10.1029/2008GL036273.
Etminan, M., G. Myhre, E. J. Highwood, and K. P. Shine, 2016: Radiative forcing of carbon dioxide, methane, and nitrous oxide: A significant revision of the methane radiative forcing. Geophys. Res. Lett., 43, 12 614–12 623, https://doi.org/10.1002/2016GL071930.
Gregory, J. M., and Coauthors, 2004: A new method for diagnosing radiative forcing and climate sensitivity. Geophys. Res. Lett., 31, L03205, https://doi.org/10.1029/2003gl018747.
Held, I. M., and B. J. Soden, 2006: Robust responses of the hydrological cycle to global warming. J. Climate, 19, 5686–5699, https://doi.org/10.1175/JCLI3990.1.
IPCC, 2021: Summary for policymakers. Climate Change 2021: The Physical Science Basis, V. Masson-Delmotte et al., Eds., Cambridge University Press, 3–32.
Joshi, M. M., J. M. Gregory, M. J. Webb, D. M. H. Sexton, and T. C. Johns, 2008: Mechanisms for the land/sea warming contrast exhibited by simulations of climate change. Climate Dyn., 30, 455–465, https://doi.org/10.1007/s00382-007-0306-1.
Kamae, Y., H. Shiogama, M. Watanabe, and M. Kimoto, 2014: Attributing the increase in Northern Hemisphere hot summers since the late 20th century. Geophys. Res. Lett., 41, 5192–5199, https://doi.org/10.1002/2014GL061062.
Knutti, R., and G. C. Hegerl, 2008: The equilibrium sensitivity of the Earth’s temperature to radiation changes. Nat. Geosci., 1, 736–743, https://doi.org/10.1038/ngeo337.
Lambert, F. H., and J. C. H. Chiang, 2007: Control of land-ocean temperature contrast by ocean heat uptake. Geophys. Res. Lett., 34, L13704, https://doi.org/10.1029/2007GL029755.
Lambert, F. H., M. J. Webb, and M. M. Joshi, 2011: The relationship between land–ocean surface temperature contrast and radiative forcing. J. Climate, 24, 3239–3256, https://doi.org/10.1175/2011JCLI3893.1.
Lu, J., and M. Cai, 2009: A new framework for isolating individual feedback processes in coupled general circulation climate models. Part I: Formulation. Climate Dyn., 32, 873–885, https://doi.org/10.1007/s00382-008-0425-3.
Manabe, S., R. J. Stouffer, M. J. Spelman, and K. Bryan, 1991: Transient responses of a coupled ocean-atmosphere model to gradual changes of atmospheric CO2. Part I. Annual mean response. J. Climate, 4, 785–818, https://doi.org/10.1175/1520-0442(1991)004<0785:TROACO>2.0.CO;2.
Sejas, S. A., O. S. Albert, M. Cai, and Y. Deng, 2014: Feedback attribution of the land-sea warming contrast in a global warming simulation of the NCAR CCSM4. Environ. Res. Lett., 9, 124005, https://doi.org/10.1088/1748-9326/9/12/124005.
Sutton, R. T., B. Dong, and J. M. Gregory, 2007: Land/sea warming ratio in response to climate change: IPCC AR4 model results and comparison with observations. Geophys. Res. Lett., 34, L02701, https://doi.org/10.1029/2006GL028164.
Toda, M., M. Watanabe, and M. Yoshimori, 2021: An energy budget framework to understand mechanisms of land–ocean warming contrast induced by increasing greenhouse gases. Part I: Near-equilibrium state. J. Climate, 34, 9279–9292, https://doi.org/10.1175/JCLI-D-21-0302.1.
Vecchi, G. A., and B. J. Soden, 2007: Global warming and the weakening of the tropical circulation. J. Climate, 20, 4316–4340, https://doi.org/10.1175/JCLI4258.1.
