Abstract
Walker circulation variability and associated zonal shifts in the heating of the tropical atmosphere have far-reaching global impacts well into high latitudes. Yet the reversed high latitude–to–Walker circulation teleconnection is not fully understood. Here, we reveal the dynamical pathways of this teleconnection across different components of the climate system using a hierarchy of climate model simulations. In the fully coupled system with ocean circulation adjustments, the Walker circulation strengthens in response to extratropical radiative cooling of either hemisphere, associated with the upwelling of colder subsurface water in the eastern equatorial Pacific. By contrast, in the absence of ocean circulation adjustments, the Walker circulation response is sensitive to the forcing hemisphere, due to the blocking effect of the northward-displaced climatological intertropical convergence zone and shortwave cloud radiative effects. Our study implies that energy biases in the extratropics can cause pronounced changes of tropical climate patterns.
INTRODUCTION
The Walker circulation is a key element of tropical climate, with air rising in the western equatorial Pacific and descending in the eastern equatorial Pacific. The Walker circulation strength is tightly coupled to the east-west contrast in equatorial Pacific sea surface temperature (SST) (1–3) through Bjerknes feedback (4). While tropical ocean-atmosphere interactions have been emphasized in the literature, recent studies hint at an extratropical impact on the equatorial Pacific SST and hence the Walker circulation. For instance, a prescribed radiative heating over the Southern Ocean (50°S to 80°S) is shown to reduce the zonal SST gradient in the equatorial Pacific (5), while a prescribed radiative cooling between 30°S and 60°S increases this gradient (6). Moreover, the leading spatial pattern of aerosol-forced SST response during the 20th century exhibits a stronger zonal SST gradient in the equatorial Pacific (7). In a future warming climate, Southern Ocean heat uptake delays extratropical Southern Hemisphere (SH) warming relative to the Northern Hemisphere (NH) counterpart, which is simulated to amplify the east-west equatorial Pacific SST gradient (8). In addition, both Arctic and Antarctic sea ice loss in the late 21st century are projected to weaken this zonal SST contrast (9, 10). Collectively, existing research highlights extratropical impacts on the spatial pattern of equatorial Pacific SST as well as its atmospheric counterpart: the tropical Walker circulation strength (11–13).
An oceanic pathway for the extratropics–to–Walker circulation teleconnection concerns the wind-driven subtropical ocean cells that connect the subduction regions in the midlatitudes with the upwelling regions in the eastern equatorial Pacific (14–17, 10). This suggests that extratropical cooling of either hemisphere acts to strengthen the Walker circulation, as colder water is upwelled in the eastern equatorial Pacific, strengthening the zonal equatorial SST gradient. Although this oceanic pathway is commonly invoked to explain the linkage between extratropical thermal perturbations and the Walker circulation, a possibility of the atmospheric pathway can also be found in the literature. For instance, slab ocean model (SOM) experiments reveal that the SH extratropical cooling displaces the intertropical convergence zone (ITCZ) northward, accompanying intensified southeasterly trade winds over the southeastern tropical Pacific where surface evaporation is enhanced to reduce SST (8, 18, 19). This wind-evaporation-SST (WES) feedback triggers the enhanced equatorial Pacific SST zonal contrast, which is further amplified by shortwave cloud radiative feedback, associated with changes of stratocumulus clouds in the southeastern subtropics (6, 8, 19). Conversely, an NH extratropical cooling, which leads to a southward ITCZ shift, has been shown to give rise to a reduced equatorial Pacific SST zonal contrast (20). Together, the Walker circulation responses to extratropical thermal perturbations seem to be dependent on the degree of atmosphere-ocean coupling (e.g., with or without ocean circulation adjustments).
Is the Walker circulation response sensitive to the hemisphere in which an extratropical energy perturbation occurs? No sensitivity is expected based on the oceanic pathway, while the atmosphere pathway would suggest a hemispheric sensitivity. The relative importance of atmospheric and oceanic pathways may vary, as extratropical energy perturbations can occur on a wide range of time scales. Our study explores the sensitivity of the Walker circulation response to the hemisphere in which an extratropical energy perturbation is imposed and the role of ocean dynamics. We perform idealized experiments in which the high latitude region of one hemisphere is radiatively cooled (fig. S1). We choose to cool, rather than to warm, the forcing region to mimic the effect of a Southern Ocean warm bias correction [i.e., climate models typically have a Southern Ocean warm bias (21)] and the effect of the NH extratropical aerosol forcing. To investigate the dynamical pathways in different components of the climate system, we use an atmospheric general circulation model coupled with a hierarchy of ocean models, ranging from a SOM on an idealized aquaplanet (AQUA) or a realistic continental configuration to a full dynamical ocean model (DOM) (Materials and Methods). The DOM experiments are part of the Extratropical-Tropical Interaction Model Intercomparison Project (ETIN-MIP) (22), allowing us to examine the robustness of the DOM experiments through the analysis of nine independent fully coupled model simulations. The SOM experiments are conducted with two independent models to examine model uncertainty. Our experimental design allows us to investigate a range of physical pathways, linking the Walker circulation response to extratropical radiative forcing in the coupled climate system.
RESULTS
Role of ocean dynamics
The longitude-latitude structure of the tropical Pacific SST response exhibits marked zonal variations in both the DOM and SOM experiments (Fig. 1), despite the zonally uniform nature of the prescribed extratropical cooling forcing. In the presence of a dynamic ocean, the east-west equatorial Pacific SST gradient is enhanced in response to extratropical radiative cooling of either hemisphere (denoted as DOM-NH and DOM-SH; Fig. 1, A and B). Consequently, the Walker circulation strengthens in the DOM experiments, inferred from multiple indices from the zonal sea level pressure gradient ∆SLP (Fig. 2A; see Materials and Methods) (23), the surface zonal wind in the central equatorial Pacific (fig. S2A) (24), to the east-west contrast of the vertical velocity in the equatorial Pacific (fig. S2B) (25). Notably, the zonal SST contrast and associated Walker circulation responses are highly robust across the ETIN-MIP ensemble (fig. S3, A and B).
Tropical SST response (shading) when (A) the NH extratropics or (B) the SH extratropics are cooled in the DOM experiments. (C and D) Same as (A) and (B) but for the SOM experiments. Hatched regions indicate statistically insignificant values at the 95% confidence level based on a two-sided t test. Black dashed lines indicate the climatological mean location of the ITCZ in the control experiment, defined as the latitude of maximum precipitation. Rectangles indicate the regions used to define the Walker circulation strength based on sea level pressure ∆SLP.
(A) The Walker circulation strength index ∆SLP response (see Materials and Methods) in the DOM and SOM experiments. The zonal gradient of equatorial Pacific SST response and its decomposition (see Materials and Methods) in the (B) DOM and (C) SOM experiments. The red bars indicate the NH extratropics cooled case, and the blue bars indicate the SH extratropics cooled case in the GFDL model, with black error bars indicating the 95% confidence level based on a two-sided t test.
From an energetics perspective, the coupled atmosphere-ocean system responds to offset the hemispheric radiative imbalance induced by the prescribed extratropical cooling (26). Both the atmosphere and ocean transport energy across the equator toward the cooled hemisphere, but meridional gradients of the respective tropical response are distinct (Fig. 3A). The atmospheric energy transport response is nearly uniform in the deep tropics, associated with small horizontal gradients of the tropical free–tropospheric temperature (27). By contrast, the oceanic response exhibits a large divergence of energy out of the tropics as a way to transport energy into the cooled extratropics. Specifically, the divergence of energy transport response by the atmosphere is less than 40% of that by the ocean in the equatorial region (fig. S4). The extratropical oceanic energy convergence (i.e., anomalously upward net surface energy flux
(A) The northward atmospheric heat transport (AHT; solid) and oceanic heat transport (OHT; dashed) response for DOM-NH (red) and DOM-SH (blue). The moist static energy transport for the atmosphere is indirectly computed from the net atmospheric column energy, while the oceanic energy transport indicates the Eulerian mean heat transport, which is a good approximation of the total ocean heat transport in the equatorial region. The circle symbol with an error bar indicates the ETIN-MIP multimodel mean and the model spread of equatorial values. (B) Same as (A) but for the SOM experiments. Blue shading in (A) and (B) indicates the forcing region. The global pattern of net surface energy flux Qnet response in (C) DOM-NH and (D) DOM-SH, with the forcing region indicated by blue lines. The Qnet response in the SOM experiments is zero by definition. Positive Qnet response indicates anomalously upward fluxes.
The effect of anomalous ocean heat uptake on enhancing the equatorial Pacific SST zonal contrast is partially cancelled by shortwave cloud radiative effects (
(A) The pressure velocity response at 500 hPa (upward in positive and downward in negative), (B) the total cloud amount response, and (C) the net shortwave cloud radiative effect (SWCRE) response over the Pacific averaged between 5°S and 5°N. Gray shaded areas indicate the regions used to define the Walker circulation strength based on sea level pressure ∆SLP. The black vertical line in (A) indicates the climatological location of maximum ascending motion in the control climate. The NH forcing case is shown in red, and the SH forcing case is shown in blue. The DOM experiments are shown in solid lines, and the SOM experiments are shown in dashed lines.
Mean ITCZ blocking effect
How then does the Walker circulation respond to extratropical radiative cooling in the absence of ocean dynamical adjustment? Note that we cannot assess the transient adjustment of the forced response from the DOM experiments (including ETIN-MIP), because the effect of internal variability cannot be removed in the presence of only one long realization. To single out the responsible atmospheric dynamics, we examine the SOM experiments in which ocean dynamics is absent, while an atmosphere-ocean thermodynamic coupling exists. The SOM experiments display a distinct zonal gradient in the equatorial Pacific SST response depending on the forcing hemisphere, with a reduced gradient (“El Niño–like” pattern) for the NH extratropical cooling case (SOM-NH; Fig. 1C) and an enhanced gradient (“La Niña–like” pattern) for the SH extratropical cooling case (SOM-SH; Fig. 1D). As a result, the Walker circulation weakens in SOM-NH while it strengthens in SOM-SH (Fig. 2A).
A closer examination of the tropical SST response pattern reveals that the SOM experiments exhibit only a weak temperature response beyond the climatological ITCZ location; that is, south of the mean ITCZ position in SOM-NH and north of the mean ITCZ position in SOM-SH (Fig. 1, C and D). The northward-displaced climatological ITCZ position seems to effectively block the NH cooling from intruding toward the equator while permitting the SH cooling to penetrate across the equator through the eastern Pacific (28). The equatorial cooling response in SOM-NH is only 40% of that in SOM-SH (Fig. 5A), whether averaged across all longitudes or over the Pacific basin (fig. S9). An energy budget analysis of the ocean mixed layer allows us to attribute the contrast in the equatorial temperature response between the two SOM experiments primarily to changes in latent heat flux related to surface wind changes (
(A) The surface temperature response averaged over 5°S to 5°N and all longitudes. (B) The attribution of equatorial surface temperature response in the SOM experiments (see Materials and Methods). The NH forcing case is shown in red, and the SH forcing case is shown in blue. Black error bars indicate the 95% confidence level based on a two-sided t test.
To further confirm the blocking effect of the mean ITCZ position, we carry out idealized AQUA experiments that are designed to isolate the mean ITCZ position effect. With a hemispherically symmetric control climate (i.e., the ITCZ is located on the equator), the climate response is insensitive to the forcing hemisphere (AQUA-SYM in Fig. 5A and fig. S12A). However, under a hemispherically asymmetric control climate (i.e., the ITCZ is displaced northward as seen in nature), the SH perturbation produces a more pronounced equatorial temperature response than the NH perturbation (AQUA-ASYM in Fig. 5A and fig. S12B), which is consistent with our conclusions drawn from the SOM experiments. The temperature response equatorward of the forcing region declines more markedly when the climatologically warmer NH is cooled, with only a very small response equatorward of the climatological ITCZ latitude, compared to when the climatologically colder SH is cooled (fig. S12B). This hemispheric sensitivity arises in response to a prescribed warming as well (fig. S13), corroborating the ITCZ blocking effect. Hence, the AQUA experiments strongly suggest that the climatological ITCZ location acts as a barrier to the propagation of the surface temperature response. This barrier is displaced north of the equator, because the NH is warmer than the SH in the Earth’s current climate (30).
With a realistic continental configuration, a clearly defined ITCZ is only present in the central to eastern Pacific (refer to Fig. 6, C and D); hence, the ITCZ blocking effect is also less prominent over the western Pacific. SOM-NH shows that a modest surface cooling response reaches the equatorial Pacific through the monsoon-dominated western Pacific warm pool region, slightly reducing the zonal contrast in equatorial Pacific SST (Fig. 1C). In comparison with SOM-NH, SOM-SH exhibits not only a substantially stronger equatorial cooling response (Fig. 5A) but also a reversed zonal contrast in equatorial Pacific SST response (Figs. 1, C and D, and 2C), as a result of the climatological ITCZ blocking effect.
(A) In the presence of ocean dynamics, the NH extratropical cooling induces colder subsurface water to be upwelled in the eastern equatorial Pacific, enhancing the zonal SST gradient and strengthening the Walker circulation. (B) Similar to (A) except for the SH forcing. (C) In the absence of ocean dynamics, the NH extratropical cooling affects the equatorial Pacific preferentially through the western Pacific warm pool associated with the climatological ITCZ blocking effect. The decreased zonal SST gradient is further reduced by the shortwave cloud radiative effects, weakening the Walker circulation. (D) Similar to (C) except for the SH forcing. The SH extratropical cooling reaches the equatorial Pacific through the eastern basin associated with climatological wind patterns via WES feedback. The increased zonal SST gradient is further enhanced by the shortwave cloud radiative effects, strengthening the Walker circulation. The horizontal map is for the SST response and the vertical cross section in (A) and (B) is for the subsurface temperature response in the equatorial Pacific. The orange contours in (C) and (D) show the 6.5 mm day−1 isopleth of the mean precipitation in the control experiment, denoting the climatological ITCZ location.
Dependence on cloud radiative effects
The strong equatorial cooling in SOM-SH is particularly pronounced off the coast of Ecuador (Fig. 1D). The southern extratropical cooling drives anomalous meridional pressure gradients that intensify the climatological southeasterlies over the eastern basins of the subtropical South Pacific (19, 31). Hence, zonal variations in climatological winds are initially responsible for shaping the spatial pattern of the equatorial Pacific cooling response through WES feedback (8, 19). Consequently, the zonal contrast in the equatorial Pacific SST response excites distinct cloud feedbacks between the eastern and western basins of equatorial Pacific. The pronounced surface cooling in the eastern equatorial Pacific in SOM-SH increases the lower tropospheric stability (fig. S14), leading to more marine low clouds and an enhanced shortwave reflection (Fig. 4, B and C, and fig. S8D) (32). Over the western equatorial Pacific, on the other hand, convective clouds shift westward associated with a strengthening of the Walker circulation, hence exhibiting a dipole pattern of shortwave cloud radiative forcing centered around 140°E, a location of the strongest deep convection in the control climate (Fig. 4). As a result, changes in shortwave cloud radiative effect in SOM-SH amplify the zonal contrast in equatorial Pacific SST (Fig. 2C). By contrast, with enhanced surface cooling over the western eastern equatorial Pacific in SOM-NH, the western Pacific warm pool deep convection shifts eastward, leading to a large negative shortwave cloud radiative effect in the central Pacific (Fig. 4). The resulting zonal variation of shortwave cloud radiative effects acts to reduce the zonal contrast in equatorial Pacific SST (Fig. 2C). Together, in the SOM experiments, changes in the shortwave cloud radiative effect ultimately control the sensitivity of the zonal SST gradient response to the forcing hemisphere (
Thus, in the absence of ocean dynamical adjustments, an intimate coupling between the SST spatial pattern, circulation, and cloud radiative effects manifests itself as the Walker circulation sensitivity to the hemisphere in which extratropical radiative cooling occurs. A crucial role played by clouds suggests a model dependence of the Walker circulation sensitivity due to large intermodel differences in cloud feedback (33). The Community Earth System Model version 1 (CESM1) exhibits a shortwave cloud radiative effect that is distinct from the Geophysical Fluid Dynamics Laboratory (GFDL) model. However, the WES feedback, with the sensitivity to the mean ITCZ position, leads to the same hemispheric sensitivity in the SST spatial pattern in the two models (fig. S3C).
DISCUSSION
It is well known that variations in the Walker circulation strength affect extratropical climate (34); however, a physical mechanism for the reverse pathway has not previously been scrutinized across different components of the climate system. The hierarchy of climate model experiments presented here reveals that the Walker circulation response via the atmospheric pathway is distinct, depending on in which hemisphere the extratropical forcing occurs, whereas the Walker circulation robustly strengthens in response to an extratropical cooling in either hemisphere when ocean dynamics is allowed to adjust (i.e., after the upper ocean reaches a quasi-equilibrium state) (Fig. 2A). The ETIN-MIP ensemble increases our confidence that the Walker circulation strengthening is a robust tropical response to an extratropical radiative cooling on multidecadal to centennial time scales. Consistent with previous studies (11, 12), upper ocean dynamics play a major role in controlling the Walker circulation response in the fully coupled climate system: Prescribed extratropical cooling lowers the temperature of water subducted within the subtropical ocean cells, causing colder water to be upwelled in the equatorial and coastal upwelling regions, thereby strengthening the zonal SST gradient and associated Walker circulation (Fig. 6, A and B).
Our hierarchical model experiments show that the atmospheric dynamics alone causes a distinct Walker circulation sensitivity to extratropical radiative forcing. The climatologically northward-displaced Pacific ITCZ blocks the NH extratropical forcing from intruding toward the equator through the central to eastern Pacific while allowing for weak surface temperature response to reach the equator over the western Pacific where deep convection is organized around the equator. By contrast, the SH extratropical forcing preferentially perturbs the surface temperatures in the eastern equatorial Pacific, because the climatological southeast trade winds allow the WES feedback from the SH. The distinct pattern of the equatorial Pacific SST response is further amplified by the shortwave cloud radiative effect, thereby creating the sensitivity of the Walker circulation response to the forcing hemisphere (Fig. 6, C and D).
One might assume that the SOM response resembles the early DOM response (say, the first few years) before the upper ocean starts adjusting to the forcing. We plan to link the SOM and DOM responses in a subsequent study by producing multiple runs with perturbed atmospheric initial conditions. Isolating the forced DOM response from the internal variability will allow us to examine the temporal evolution in the initial adjustment.
Our results suggest that extratropical climate change may be an important factor for modulating the Walker circulation. The extratropical influence may explain the recent strengthening trend of the Walker circulation in observations (24). For example, anthropogenic aerosols that have cooled the northern extratropics throughout the 20th century may have contributed to the observed Walker circulation strengthening. Our results imply that reducing uncertainty in simulated extratropical climate may be an important step forward for improving model simulations of tropical SST patterns and the associated Walker circulation as well as their future projections in response to radiative forcing.
MATERIALS AND METHODS
Model and experiment description
We use a developmental version of the GFDL next-generation atmospheric model version 4 (AM4) (35) coupled with (i) a DOM used in the GFDL Forecast-Oriented Low Ocean Resolution (36) under a preindustrial condition or (ii) a SOM with prescribed climatological surface heat fluxes (often referred to as Q-flux) to mimic climatological mean SSTs simulated by the DOM control experiment. Both the atmosphere and ocean models have an approximate 1° horizontal resolution. The atmospheric model has 32 vertical levels, and the ocean model has 50 vertical levels. The preindustrial control climate is perturbed by reducing solar flux over either 45°S to 65°S or 45°N to 65°N to produce a total energy perturbation of about 0.8 PW in each experiment (fig. S1). The maximum forcing magnitude of 46 W m−2 at 55°S/N is comparable to the intermodel spread of shortwave cloud radiative forcing over the similar latitude band in the SH (37). This applied radiative forcing is zonally uniform and identical except for the hemisphere that it is applied to. The NH extratropical forcing broadly corresponds to the regions of large anthropogenic aerosol emissions and/or continental ice sheets that wax and wane during glacial-interglacial cycles, while the SH forcing location corresponds to the region of pronounced cloud biases with the largest intermodel spread (37). These experiments can also be interpreted within the context of potential future geoengineering impacts. The DOM experiments are initialized from a fully spun-up preindustrial control run and are integrated for 150 years for both the control and perturbation experiments. The analysis is based on the average of years 101 to 150, considering that the simulations reach a quasi-equilibrium state for the upper ocean around year 70. All SOM experiments are integrated for 100 years, and the last 40 years are used for the analysis. The experiment setup in this study is the protocol used for ETIN-MIP (22) so that the results from nine independent fully coupled models are overlaid with the DOM experiment results in Fig. 3A and figs. S3 and S9 to confirm the robustness of the results. To address the uncertainty associated with the SOM response, the SOM experiment is also conducted with the Community Earth System Model version 1.2.2 (CESM1) (38), as shown in figs. S3 and S9.
In addition, the AM2 model coupled to an aquaplanet SOM (AQUA) (39) is used for experiments in which the degree of hemispheric asymmetry in the control climate is altered. The AQUA-ASYM experiment is forced with the zonal mean of observed surface heat flux (solid line in fig. S15) so that the NH is simulated to be warmer than the SH as in the current climate. The AQUA-SYM climate state is achieved by symmetrizing the observed surface heat flux between hemispheres (dashed line in fig. S15). The control climate with a varying degree of hemispheric asymmetry is then perturbed by adding an anomalous surface heat flux that has the same profile as the top-of-atmosphere radiative perturbations in the DOM and SOM experiments. We also examine the sensitivity to the sign of surface flux perturbations (fig. S13).
Walker circulation strength index
Following (23), we define the Walker circulation strength index (denoted as ∆SLP) as the sea level pressure difference over the central/east Pacific (EP; 160°W to 80°W, 5°S to 5°N) and over the Indian Ocean/west Pacific (WP; 80°E to 160°E, 5°S to 5°N; see Fig. 1). Since changes in the Walker circulation strength occur in conjunction with changes in the zonal gradient of equatorial Pacific SST, we relate ∆SLP to the surface temperature difference over the Indian Ocean/WP and over the central/EP in Fig. 2.
Surface energy budget
We diagnose the drivers of the equatorial surface temperature changes using a surface energy budget decomposition. Motivated by Hwang et al. (8), the surface temperature response is attributed to the following factors: changes in ocean heat uptake (
Following the bulk formula for evaporation, changes in the latent heat flux related with Newtonian cooling can be expressed as
We emphasize that
SUPPLEMENTARY MATERIALS
Supplementary material for this article is available at http://advances.sciencemag.org/cgi/content/full/6/47/eabd3021/DC1
This is an open-access article distributed under the terms of the Creative Commons Attribution-NonCommercial license, which permits use, distribution, and reproduction in any medium, so long as the resultant use is not for commercial advantage and provided the original work is properly cited.
REFERENCES AND NOTES
- Copyright © 2020 The Authors, some rights reserved; exclusive licensee American Association for the Advancement of Science. No claim to original U.S. Government Works. Distributed under a Creative Commons Attribution NonCommercial License 4.0 (CC BY-NC).