Model consistency for the underlying mechanisms for the Inter-decadal Pacific Oscillation-tropical Atlantic connection

Modeling evidence suggests that the Inter-decadal Pacific Oscillation (IPO) can remotely affect the tropical Atlantic sea surface temperature (SST) variability. However, the root causes of the IPO-tropical Atlantic inter-basin teleconnections are not fully understood. Using idealized pacemaker experiments wherein the observed IPO-SST anomalies are specified, we show that a warm-phase IPO-SST anomaly drives a basin-wide SST warming over the tropical Atlantic. The trans-basin IPO-tropical Atlantic connection is established via the tropical tropospheric temperature mechanism and the atmospheric bridge teleconnections. An IPO positive-phase SST initiates the warming tropospheric temperature anomalies, leading to a more humid atmosphere and increasing longwave radiation downward into the tropical Atlantic. By comparison, the reduced Pacific Walker circulation and mid-latitude Rossby wave responses cause the trade winds to strengthen, acting to suppress the tropical Atlantic warming through the Bjerknes positive feedback mechanism. Thus, similar to previous studies, the tropical Atlantic widespread warming is ultimately traced back to the warming effects generated by the water vapor-longwave radiation-SST positive feedback, which, contrary to previous studies, overwhelms the cooling effects associated with the intensified trade winds. Our results highlight the model dependence on the details of mechanisms that connect the tropical Pacific and tropical Atlantic on the decadal timescales.


Introduction
Modeling and observational evidence suggest that the Pacific sea surface temperature (SST) variations substantially influence the tropical Atlantic Ocean, affecting its inter-annual and decadal timescale variability (Chang et al 2006b, Cai et al 2019. The Atlantic SST variability modulates the global climate system, with widespread climatic consequences. For example, anomalous warming over the north tropical Atlantic drives a northward swing of the Atlantic intertropical convergence zone (Chang et al 2000), thus causing droughts in Northeast Brazil and centraleastern Amazon (Hastenrath and Heller 1977), as well as floods across northern South America and Caribbean regions (Grimm and Tedeschi 2009). The north tropical Atlantic warming also induces increasing risks of landfall hurricane frequency in the U.S. (Xie et al 2005), triggers the central Pacific La Niña events (Ham et al 2013), and influences the Antarctic sea ice redistribution (Li et al 2014(Li et al , 2021. Thus, determining the potential mechanisms for the trans-basin Pacific-Atlantic connections is societally essential. The tropical Pacific-Atlantic teleconnections have been well established on the inter-annual timescale. The El Niño-Southern Oscillation (ENSO), the most noticeable year-to-year variation on Earth, originates in the tropical Pacific and alternates between warm El Niño and cold La Niña events. After an El Niño matures, the north tropical Atlantic typically warms for a few months during the boreal spring and early summer (Chiang and Sobel 2002, Kumar and Hoerling 2003, Taschetto et al 2016 as a consequence of reduced surface latent heat flux associated with the weakened northeasterly winds via the atmospheric bridge teleconnections (Klein et al 1999) or the dampened moist convection and circulation via the tropospheric temperature mechanism (Chiang and Sobel 2002). The Pacific El Niño-like warming also impacts the equatorial and south tropical Atlantic, and the associated responses are relatively weak. Specifically, an El Niño event produces a reduced Pacific Walker circulation and strengthened trade winds in and to the south of the equatorial Atlantic (Chiang andSobel 2002, Sasaki et al 2015), inducing cold SST anomalies along the equator and to the south of the tropical Atlantic during the boreal spring and summer through Bjerknes feedback (Bjerknes 1969) that may counter the tropospheric temperature warming response (Chang et al 2006a). Thus, warm or neutral conditions may occur in the equatorial and south tropical Atlantic.
On the decadal timescales, the physical processes for the trans-basin Pacific-Atlantic teleconnections remain controversial (Latif 2001). The Interdecadal Pacific Oscillation (IPO), sometimes also called Pacific Decadal Oscillation (Power et al 1999, Newman et al 2016, emerges as the leading Empirical Orthogonal Function mode of Pacific decadal SST variability after the radiatively forced component has been subtracted at each grid point (Zhang et al 1997). Observations document the interbasin Pacific-Atlantic links in decadal SST variability (Nigam et al 2020), with the warm tropical Pacific leading with the same sign in the tropical Atlantic by ∼three years (Meehl et al 2021), despite the lack of the involved mechanisms due to the limited instrument records. Modeling studies show that a warm-polarity IPO SST forces anomalous warming in the tropical Atlantic through the Walker circulation perturbations and mid-latitude atmospheric Rossby wave responses (Meehl et al 2021). However, little is known about whether the proposed mechanisms are model-dependent. Less appreciated is which underlying mechanism is largely responsible for the IPOtropical Atlantic connections.
Here, we revisit a recent study of CESM1 (Meehl et al 2021) and analyze similar idealized Pacific pacemaker experiments with two additional coupled models to investigate the model dependence on the tropical Pacific-Atlantic connections. Our study confirms that an IPO warm-phase SST drives the basin-wide decadal SST warming in the tropical Atlantic via a tropical troposphere temperatureinduced surface heat flux mechanism. We also demonstrate for these two models that the proposed mechanism overwhelms latent heat flux reduction related to the enhanced trade winds over the tropical Atlantic.

Methods and experimental setup
To illuminate how a warm-phase IPO-SST anomaly affects the tropical Atlantic climate anomalies, we evaluate idealized IPO pacemaker simulations with only two current coupled models, including IPSL-CM6A-LR and HadGEM3-GC31-MM, wherein a time-invariant spatial pattern corresponding to the internally-generated component of observed IPO-SST anomalies is constrained, as a contribution to the Decadal Climate Prediction Project (Boer et al 2016). The atmospheric component of IPSL-CM6A-LR has 144 × 143 longitude/latitude and 79 vertical layers with a ∼80 km top-level. Its ocean model has a quasi-isotropic tripolar grid of 1 • nominal resolution and 75 ocean levels (Boucher et al 2020). HadGEM3-GC31-MM has a high-resolution atmospheric component of 432 × 324 longitude/latitude and 85 vertical layers with a ∼85 km top-level, together with an oceanic component having a tripolar grid of ∼0.25 • and 75 • ocean levels (Roberts et al 2019). Both models reasonably reproduce the characteristics of ENSO and IPO, despite common mean-state SST biases (Roberts et al 2019, Boucher et al 2020. The IPO index and its SST pattern utilized in idealized pacemaker simulations are calculated from the observed SST from 1900 to 2013 (ERSST V4). Following earlier approaches (Ting et al 2009, Ruprich-Robert et al 2017, Meehl et al 2021, Yao et al 2021, the externally-forced SST response is first taken from the CMIP5 multi-model ensemble under historical simulations  and Representative Concentration Pathway 8.5 (RCP8.5;2006 and then removed from observations. Finally, the IPO index is estimated as the leading principal component of an empirical orthogonal function decomposition to the observed residual SST over the Pacific domain (40 • S-60 • N, from Indonesia to the American Coast). The IPO-SST spatial pattern is obtained by regressing the IPO index onto SSTs at each grid point from 1900 to 2013 (figure S1).
The Pacific-wide SSTs are restored to an IPO-SST spatial pattern in these two global coupled models by adding the non-solar surface heat fluxes to the mixed layer (Haney 1971, Frankignoul and Kestenare 2002, Servonnat et al 2015. Here, the heat flux adjustments are made following the restoring term: where dQ dT is a constant restoring coefficient of −40 W m −2 K −1 corresponding to a restoring timescale of approximately 60 days for a 50 mdepth mixed layer. Outside the targeted region, SSTs are free to evolve and are coupled to the overlying atmosphere. Table S1 outlines two sets of 10-member and 25-member ensemble experiments of IPSL-CM6A-LR and HadGEM3-GC31-MM for the idealized IPO pacemaker configurations. In the full IPO pacemaker simulations, an IPO-SST anomaly pattern of positive-polarity and negative-polarity is separately prescribed over the Pacific domain (40 • S-60 • N and from Indonesia to the American Coast in figure  S1(a)), hereafter denoted as IPO+ and IPO−. In the North IPO simulations (called Nor_IPO+ and Nor_IPO−), similar to the IPO+ and IPO− but only the northern part of IPO is specified from 20 • N to 60 • N (figure S1(b)), which approximately represents a Pacific Decadal Oscillation (PDO)-SST anomaly (Newman et al 2016). These two distinctive pacemaker experiments aim to gain insight into the relative importance of IPO and PDO in driving the transbasin Pacific-Atlantic teleconnections. All radiative forcing agents are kept constant at the pre-industrial levels in each experiment. Each realization is run for only ten-model years to avoid creating climate anomalies over the unrestored subsurface ocean caused by deep ocean circulation that could deteriorate the tropical Atlantic climate response to the IPO SST forcing. We assume the linearity of the Atlantic climate response to the warm-phase and cold-phase IPO and directly examine the consequence of a negative-topositive phase transition of IPO (disregarding a neutral state). To maximize the signal-to-noise ratio, we extract the Atlantic climate response to a warm-phase IPO-SST forcing by taking the difference between IPO+ and IPO− ensemble simulations rather than comparing the perturbation experiments to a control run. The results presented here are calculated from the ensemble mean of each model experiment. In this study, we analyze each year of the individual ensemble realizations to evaluate the statistical significance of the difference between IPO+ (North_IPO+) and IPO− (North_IPO−) using the two-sided Student t-test. Figure 1 exhibits the Atlantic decadal-timescale climate response to the full and North IPO-SST anomalies. For the full IPO warm-polarity case (IPO+ minus IPO−), most of the tropical Pacific is characterized by significant warm SST anomalies of over 0.4 • C (figures 1(a) and (b)), comparable to the observed SST (figure S1). Both models consistently simulate the widespread warm SST anomalies over the tropical Atlantic, with a stronger magnitude in

The potential impact of an IPO-SST anomaly on the Atlantic climate change
. Correspondingly, anomalous westerly surface winds appear over the centralwestern equatorial Pacific, while easterly surface wind anomalies occupy the far-eastern Pacific, tropical Atlantic, and the equatorial Indian Oceans. Alternating anomalous cyclones and anti-cyclones extend from the North Pacific to North America and South Pacific to the Amundsen Sea, ending with anomalous anti-cyclones centered over the North and South tropical Atlantic, respectively. Unlike the full IPO-SST forcing (figures 1(a) and (b)), a North IPO-SST warming, representing a warm-phase PDO, forces relatively weak warm SST anomalies and reduced trade winds over the tropical Atlantic (figures 1(c) and (d)). We note that an anomalous anti-cyclone is seen over the North Atlantic, in contrast to an anomalous cyclone resulting from a full IPO warm-phase SST forcing. The results suggest that a full IPO warmphase SST, particularly in the tropical domain that plays a major role (Deser et al 2004), significantly drives the decadal-scale climate anomalies in the tropical Atlantic.
Although the IPO and PDO indices are highly correlated (r = ∼0.9 with >95% significance) in observations (Han et al 2014), their respective global influences in coupled models are distinctive. Compared to a cold-phase PDO (Nor_IPO− minus Nor_IPO+ in figure S2), an IPO-SST cooling forces stronger surface easterly wind anomalies over the tropical Pacific and larger cold SST anomalies worldwide, supporting the notion of a cold-phase IPO dominating the early-2000s hiatus (Kosaka andXie 2013, Meehl et al 2016). As corroborated by recent findings (Meehl et al 2021), an IPO-SST anomaly triggers a basin-wide SST response across the tropical Atlantic on the decadal timescale, with comparable magnitude in the same sign as the tropical Pacific. In contrast, a warm tropical North Atlantic typically follows an El Niño event on the interannual timescale (Curtis and Hastenrath 1995). Therefore, we aim to provide helpful insight into the potential mechanisms for a warm-phase IPO-SST anomaly exciting the tropical Atlantic basin-scale warming.

Tropical and extra-tropical dynamical pathways for affecting the tropical Atlantic climate response
We explore the respective roles of the tropical and extra-tropical atmospheric teleconnections in relaying the warm-polarity IPO-SST signal to the tropical Atlantic. The atmospheric baroclinic vertical structure in the tropics is estimated by half the difference between stream functions at 850 hPa and 200 hPa levels (Yu et al 2010). An IPO-SST warming drives a southward shift of the North Pacific intertropical convergence zone and a northeastward shift of the South Pacific convergence zone (figures 2(a) and (b)), accompanied by an enhanced ascending motion over the equatorial eastern Pacific and a reduced Pacific Walker circulation (figures 2(c) and (d)). The associated convective heating anomalies in the tropical eastern Pacific excite a baroclinic atmospheric response within the tropics, characterized by a pair of baroclinic cyclones straddling the equatorial Pacific and another pair of baroclinic anti-cyclones covering the tropical Atlantic and tropical Indian Oceans. This circulation pattern resembles a typical Gill-Matsuno response (Gill 1980), as expected from the composite four-season evolutions of ten-year idealized pacemaker simulations in both models (figures S3 and S4). In the January-February-March, the eastern Pacific warming-driven atmospheric deep convection initiates an equatorial Kelvin wave, generating easterly surface wind anomalies across the equatorial Atlantic and Indian Oceans, as well as Rossby wave packets associated with equatorial westerly wind anomalies over the western Pacific. From April-May-June to July-August-September, the eastern Pacific convection activity enhances and shifts westward toward the International Dateline. By October-November-December, the prescribed warm-phase IPO SST anomaly eventually coexists with the basinwide warm SST in the tropical Atlantic.
Additionally, in response to a warm-polarity IPO SST anomaly, significant tropospheric temperature warming occurs the entire tropical free troposphere (figures 2(e) and (f)), as measured by the column-integrated tropospheric atmospheric temperature anomalies from 1000 hPa to 200 hPa. More overlooked, however, is that the tropospheric temperature significant warming in the tropical eastern Pacific is almost in phase with coherent warming across the tropical Atlantic and Indian oceans. By comparing the composited four-season evolutions of column-integrated tropospheric temperature anomalies from the ten-year idealized pacemaker simulations (figure S5), it is clear that the magnitude of the tropospheric temperature response over the tropical Atlantic and Indian Oceans is more dependent on the tropospheric atmospheric temperature warming anomalies over the tropical eastern Pacific. These results identify the predominance of an IPO warm-phase SST over the tropospheric atmospheric temperature warming on the decadal timescales, in close agreement with earlier findings of the ENSO mediating the inter-annual tropospheric temperature variability across the tropical Oceans (Pan and Oort 1983, Yulaeva and Wallace 1994, Soden 2000. In addition to the tropical atmospheric teleconnections, a positive-phase IPO-SST anomaly can also remotely influence the North and South Atlantic via the mid-latitude atmospheric teleconnections. An IPO-SST warming yields positive precipitation anomalies and the enhanced latent heat release over the tropical eastern Pacific (figures 2(a) and (b)), subsequently heating the overlying atmosphere and triggering a two-pronged, large-scale Rossby wave train (figures 3(a) and (b)). This Rossby wave response first extends from the tropical Pacific to the North and South Pacific, then propagates eastward, and ultimately reaches the North and South Atlantic, respectively. The composite seasonal evolutions of the wave activity flux for the ten-year idealized pacemaker simulations provide strong support for the midlatitude Rossby wave response (Takaya and Nakamura 2001) from 30 • N to 80 • N in the North Hemisphere and from 30 • S to 80 • S in the South Hemisphere (figure S6). These alternating active low-pressure and high-pressure centers bear a striking resemblance to an anomalous Pacific-North America pattern to the north and east (Wallace and Gutzler 1981) and an anomalous Pacific-South America Pattern to the south and east (Mo 2000), each with highly equivalent barotropic structures with amplitude from sea level to mid-troposphere (500 hPa) and even up to 200 hPa ending with low center anomalies in the extra-tropical North and South Atlantic ( figure 3).
As a result, increased sea level pressure is seen over the tropical Atlantic (figures 3(e) and (f)), inducing an enhanced zonal pressure gradient between the tropical Atlantic-tropical eastern Pacific and thus favoring intensified trade winds over the tropical Atlantic. The strengthened easterly winds are conducive to SST cooling over the tropical Atlantic via the Bjerknes feedback mechanism (Bjerknes 1969), in contrast to the tropospheric temperature warming mechanism. Note that the tropical Atlantic easterly wind strength relies largely on the location of lowpressure anomalies over the extra-tropical Atlantic. To better understand the role of a warm-phase IPO-SST anomaly in determining the trans-basin Pacific-Atlantic connection, we further evaluate the contributions of a PDO positive-phase SST anomaly to generate tropical and mid-latitude atmospheric teleconnections. We found that a PDO-SST forcing exerts a marginal impact on the tropospheric atmospheric temperature warming (figure S7) and midlatitude Rossby wave responses ( figure S8). The distinctive tropical Atlantic decadal SST responses to the warm-polarity IPO-SST and PDO-SST anomalies in both models emphasize the importance of an IPO-SST warming in triggering the tropical Pacifictropical Atlantic decadal teleconnections.

Potential mechanisms for a warm-phase IPO-SST driving the tropical Atlantic warming
A positive-polarity IPO-SST anomaly linkage to the tropical Atlantic on the decadal timescale has been successfully established via the tropical and midlatitude atmospheric teleconnections in two global coupled models and discussed in the context of earlier comparable results from a third coupled model (Meehl et al 2021). We investigate in more detail the major driver for a warm-phase IPO-SST anomaly leading to basin-wide decadal warming in the tropical Atlantic. Following earlier studies (Xie et al 2010, Yao et al 2021, the ocean mixed-layer heat-budget analysis is expressed as: where T, c o p , ρ o , and H are SST, the specific heat at constant pressure, the density of seawater, and the mixed-layer depth (a constant 50 m), respectively. The heat-budget terms on the right-hand side of the equation encapsulate longwave radiation (Q L ), shortwave radiation (Q S ), latent heat flux due to changes in near-surface relative humidity ( QE e α∆T −RH RH ′ ), changes in surface atmospheric stability ( αQERH e α∆T −RH ∆T ′ ), changes in near-surface wind speed ( QĒ W W ′ ), and sensible heat flux (Q H ). D o represents the sum of zonal, meridional, and vertical advection of ocean heat transport. The residual term consists of heat transport related to unresolved processes, including diffusion, submesoscale processes, and mixing and diabatic processes (Abellán et al 2018). Our study concerns the SST spatial pattern formation in the tropical Atlantic for the ten-year idealized pacemaker simulations (IPO+ minus IPO−). Because all radiative forcings are kept constant at the pre-industrial levels in both models, the SST change ( ∂T ∂t ) is in a quasi-equilibrium between the four surface heat flux components and ocean dynamics (D o ) (see a complete analysis of the ocean mixedlayer heat budget analysis shown in Text S1 in the supplementary information). As ocean current velocities are not currently available in the IPO idealized pacemaker simulations for some realizations of HadGEM3-GC31-MM, the presentation of our main results is centered on IPSL-CM6A-LR. Given a robust validation of the IPO-tropical Atlantic connections in both climate models, our main conclusions hold for HadGEM3-GC31-MM.
We found that the downward longwave radiation increase is the primary contributor to relaying the IPO-SST warming signal to the basin-wide SST over the tropical Atlantic ( figure 4(a)), in line with earlier results (Meehl et al 2021). Although other surface heat flux components, including shortwave radiation, sensible heat flux, and latent heat flux, along with ocean dynamics, contribute partly to SST warming at some locations in the tropical Atlantic sector (figure S9), all of them are not sufficient to fully explain the basin-scale warming in the tropical Atlantic. More specifically, a warm-phase IPO-SST anomaly heats the tropical tropospheric atmospheric temperature aloft over the Atlantic via the eastward-propagating Kelvin wave (Yulaeva and Wallace 1994, Chiang andSobel 2002). Moreover, the altered Walker circulation induces anomalous subsidence in the tropical Atlantic, adiabatically warming the tropospheric atmosphere. Both atmospheric processes allow the tropospheric atmosphere to hold more moisture ( figure 4(b)) and increase the total cloud fraction associated with larger relative humidity (solid contours in figure 4(d)). Because atmospheric water vapor absorbs infrared radiation and is an important greenhouse gas, a larger concentration of water vapor results in higher SSTs by increasing downward longwave radiation (figure 4(a)). As SST increases, more water vapor from the warm ocean will evaporate into the overlying atmosphere, and a positive feedback mechanism between water vapor and SST is established (Chu and Murakami 2022). Thus, the tropical Atlantic widespread SST warming is dominated by the downward longwave radiation, whose contribution is larger than that due to an increase in the relative humidity-induced and atmospheric stability-related latent heat flux (figures 4(d) and (e)).
These results disagree with previous findings, which suggest that the latent heat flux and solar radiation give rise to the tropical north Atlantic warming response to ENSO on the inter-annual timescale (Curtis and Hastenrath 1995, Klein et al 1999, Chiang and Sobel 2002. Our study highlights that the warming effects induced by the water vaporlongwave radiation-SST positive feedback completely (a) Longwave radiation, (b) atmospheric water vapor content, (c) total cloud fraction, and latent heat flux due to (d) variations in relative humidity (shading denotes relative humidity-related latent heat and contours with an interval of 0.1% from −0.4% to 0.4%; solid contours represent increased relative humidity), (e) changes in atmospheric stability (shading represents stability-related latent heat and contours with an interval of 0.01 • C from −0.04 • C to 0.04 • C; solid contours indicates enhanced air-sea temperature difference), (f) changes in near-surface wind speed (shading represents wind speed related latent heat and contours with an interval of 0.01 m s −1 from −0.08 m s −1 to 0.08 m s −1 ; solid contours denote increased wind speed). The surface heat flux is positive downward. Stippling denotes regions exceeding 95% statistical significance based on a two-sided student t-test.
negate the cooling effects arising from the increased latent heat loss due to strengthened near-surface speed via both the Walker circulation weakening and mid-latitude Rossby wave response (figure 4(f)), consistent with earlier studies (Meehl et al 2021) indicating that the teleconnections produced the tropical Atlantic weakened easterlies and latent heat flux reduction that added to the warming effects from increased downward longwave radiation. Our results further confirm that the water vapor-longwave radiation-SST positive feedback should be mainly responsible for the decadal-timescale tropical Atlantic basin-wide warming response to a warm-phase IPO-SST anomaly.

Conclusion
Using a series of idealized warm-phase and coldphase IPO pacemaker experiments with two global coupled models, we have shown that a warm-phase (cold-phase) IPO-SST anomaly drives basin-wide SST warming (cooling) response over the tropical Atlantic through both the tropical and mid-latitude atmospheric teleconnections. For a warm-polarity IPO-SST case, the tropical tropospheric atmosphere markedly warms and can thus hold more water vapor. A warmer and more damp tropospheric atmosphere causes downward longwave radiation to increase, eventually heating the basin-wide SST across the tropical Atlantic. On the contrary, a weakened Pacific Walker circulation, in combination with the mid-latitude Rossby wave responses, amplifies the sea level pressure over the tropical Atlantic. The resultant tropical Atlantic-Pacific trans-basin pressure gradient increases and produces strengthened trade winds over the tropical Atlantic, especially in IPSL-CM6A-LR, contrary to HadGEM3-GC31-MM and CESM1 (Meehl et al 2021). The trade wind intensification seems to cool the tropical Atlantic SST via the Bjerknes positive feedback mechanism. Accordingly, the large-scale warming in the tropical Atlantic, as a consequence of a warm-phase IPO-SST forcing, is dominated by the tropospheric atmospheric temperature mechanism that overwhelmingly counterbalances the atmospheric bridge mechanism comprising a reduced Pacific Walker circulation and the mid-latitude Rossby wave responses.
A possible caveat is that the IPO-tropical Atlantic decadal teleconnections may be model-dependent. Notably, the tropical Atlantic trade winds depend more on the geographic pattern of mid-latitude teleconnections. For IPSL-CM6A-LR, anomalous lowpressure centers of action in the extra-tropical North and South Atlantic reside near 50 • N and 65 • S (figure 3(e)), too farther poleward to impact the tropical Atlantic trade winds, leading to the prevailing easterlies between 20 • N and 20 • S ( figure 1(a)). By comparison, anomalous low-pressure active centers dwell near 30 • N and 50 • S in HadGEM3-GC31-MM (figure 3(f)) and CESM1 (Meehl et al 2021), and comparatively, weak trade winds appear only along the equatorial Atlantic ( figure 1(b)). Overall, large increases in downward longwave radiation effectively negate latent heat flux loss related to the strengthened trade winds, eventually leading to a basin-wide SST warming over the tropical Atlantic. The three coupled model results reflect a physical consistency of the tropical Atlantic basin-scale warming response to a warm-polarity IPO-SST forcing.
Our results suggest that an IPO-SST anomaly creates a same-sign SST response in the tropical Atlantic. However, the converse is not the case. Previous studies have documented that the Atlantic SST rapid warming during a warm-polarity Atlantic Multi-decadal Variability (AMV) amplifies cold SST anomalies over the eastern-central tropical Pacific (Kucharski et al 2011, McGregor et al 2014, Li et al 2016, Ruprich-Robert et al 2017, Meehl et al 2021, Yao et al 2021. On the other hand, recent modeling studies argue that while the IPO mostly originates from internal variability, external forcings contribute to the tropical Pacific decadal SST variations since the 1990s, and anthropogenic aerosols have modulated the magnitude of a cold-phase IPO-SST anomaly since 1998 (Hua et al 2018). The AMV is, instead, a consequence of the combination of internal variability and volcanic and anthropogenic aerosols (Qin et al 2020a(Qin et al , 2020b. Thus, it remains challenging to determine what leads the IPO to change phase and what causes the AMV itself to shift phase. However, at least to a certain degree, we realize that the Pacific and Atlantic could be mutually interactive on the decadal timescale (e.g. about ten years), with one alternatively impacting another (Meehl et al 2021). A broader understanding of the decadal-scale trans-basin Pacific-Atlantic climate interactions may provide a physical basis for improving decadal predictions of global warming rates, considering emerging evidence that the net impact of multiple ocean basin SST changes contributes to the global warming acceleration and slowdown (Drijfhout et al 2014, Yao et al 2017.

Data availability statement
All the model data presented in our study are publicly available at https://catalogue.ceda.acuk/uuid/ 4bf2f8d1dfd6470ba783f36bc9f581b3?jump=relatedanchor