Climate Dynamics

, Volume 40, Issue 5, pp 1435–1452

Global energy budget changes to high latitude North Atlantic cooling and the tropical ITCZ response


    • Centre for Ice and Climate, Niels Bohr InstituteUniversity of Copenhagen
    • Earth and Planetary ScienceUniversity of California
  • John C. H. Chiang
    • Department of Geography, Berkeley Atmospheric Sciences CenterUniversity of California

DOI: 10.1007/s00382-012-1482-1

Cite this article as:
Cvijanovic, I. & Chiang, J.C.H. Clim Dyn (2013) 40: 1435. doi:10.1007/s00382-012-1482-1


The intertropical convergence zone (ITCZ) in atmospheric general circulation models (coupled to slab ocean) shift southwards in response to northern extratropical cooling. Previous studies have demonstrated the utility of diagnosing the atmospheric energy fluxes in interpreting this teleconnection. This study investigates the nature of global energy flux changes in response to North Atlantic high latitude cooling applied to the Community Atmosphere Model version 3 coupled to a slab ocean, focusing on key local and remote feedbacks that collectively act to alter the energy budget and atmospheric energy transport. We also investigate the relative roles of tropical sea surface temperature (SST) and energy flux changes in the ITCZ response to North Atlantic cooling. Using a radiative kernel technique, we quantify the effects of key feedbacks—temperature, cloud and water vapor, to the top-of-the-atmosphere radiative flux changes. The results show only partial local energy flux compensation to the initial perturbation in the high latitudes, originating from the negative temperature feedback and opposed by positive shortwave albedo and longwave water vapor feedbacks. Thus, an increase in the atmospheric energy transport to the Northern extratropics is required to close the energy budget. The additional energy flux providing this increase comes from top-of-the-atmosphere radiative flux increase over the southern tropics, primarily from cloud, temperature and longwave water vapor feedbacks, and largely as a consequence of increased deep convection. It has been previously argued that the role of tropical SST changes was secondary to the role played by the atmospheric energy flux requirements in controlling the ITCZ shifts, proposing that the SST response is a result of the surface energy budget and not a driver of the precipitation response. Using a set of idealized simulations with the fixed tropical SSTs, we demonstrate that the ITCZ shifts are not possible without the tropical SST changes and suggest that the tropical SSTs are a more suitable driver of tropical precipitation shifts compared to the atmospheric energy fluxes. In our simulations, the ITCZ shifts are influenced mainly by the local (tropical) SST forcing, apparently independent of the actual high latitude energy demand.


ITCZ shiftTropical sea surface temperaturesEnergy budgetRadiative feedbacks

1 Introduction

Tropical paleoproxies record a number of rapid shifts in the hydrologic cycle across the last glacial (Lea et al. 2003, 2006; Fleitmann et al. 2003; Peterson and Haug 2006; Kienast et al. 2006; Dubois et al. 2011; and many others). Moreover, a number of studies propose the tropical changes (precipitation or sea surface temperature) being in phase with Greenland temperature changes (Peterson et al. 2000; Kienast et al. 2001; Peterson and Haug 2006; Koutavas and Sachs 2008; Wang et al. 2008), thus implying the existence of a robust teleconnection between the high northern latitudes and tropics.

High northern latitude temperature perturbations simulated in atmospheric or coupled models have been shown to shift the meridional position of the Intertropical Convergence Zone (ITCZ) to the anomalously warmer (southern) hemisphere (Chiang and Bitz 2005; Broccoli et al. 2006; Chiang et al. 2008; Mahajan et al. 2011). The shift in the ITCZ implies a meridional displacement of the rising branch of the Hadley circulation. Viewed in terms of the angular momentum conservation, the circulation shifts towards the warmer hemisphere in order to avoid the high temperature gradients in the upper part of the Hadley cell (due to asymmetric cooling/heating in the subtropics) as shown by Lindzen and Hou (1988).

Two perspectives of tropical ITCZ shifts to high latitude cooling had emerged. The first view, which will be referred to as the SST perspective, argues that the tropical interhemispheric SST gradient has to be altered in order for the ITCZ to respond, in accordance with a longstanding view that the ITCZ position is controlled by cross-equatorial SST gradients. Such interpretation is based on an argument by Lindzen and Nigam (1987) that tropical surface pressure gradients are tied to SST gradients through the effect that the latter have on temperatures in the boundary layer. This is because the temperature anomalies in the tropical boundary layer are essentially bound to the underlying SST anomaly. Thus, if one further assumes that pressure gradients above the boundary layer are negligible (‘weak temperature gradient’ assumption), it follows that the surface pressure gradients are determined by the SST gradients. Anomalous cross-equatorial boundary layer flows are therefore driven by interhemispheric SST gradients, and hence the ITCZ position. This basic relationship between the tropical interhemispheric SST gradient and the ITCZ forms the much of the basis in interpreting modern-day tropical Atlantic interannual-decadal variability, whose dominant pattern of SST-precipitation change is exactly of this type (e.g. Xie and Carton 2004). Similar conclusions have been derived in tropical Atlantic coupled ocean–atmosphere studies by Chang et al. (2000) and Chiang et al. (2002). Following this view, Chiang and Bitz (2005) proposed a mechanism for the communication of the high latitude Northern Hemisphere cooling to the tropics through the progression of the surface cooling by the equatorward advection of cold air in the midlatitudes, and a wind-evaporation-sea surface temperature (WES) feedback in the subtropics (Xie 1999). As the cooling reaches the tropical latitudes, anomalous cooling in the Northern subtropics creates the anomalous meridional pressure gradient that drives an anomalous southerly cross-equatorial flow and causes the ITCZ displacement (Chiang and Bitz 2005). Positive feedbacks may act to augment the existing meridional SST gradient and cross-equatorial wind anomalies, including a ‘near-equatorial’ WES effect and low cloud feedbacks (Xie and Carton 2004).

From the energetics viewpoint, cooling in the high northern latitudes is only partially compensated by local (i.e. high latitude) radiative flux changes, and an increased northward atmospheric heat transport is required to close the extratropical energy budget (Chiang and Bitz 2005). The additional energy is provided by the southern tropics, from increased radiative flux absorption resulting from the ITCZ shift to the anomalously warmer hemisphere. This energy surplus is then transported from the southern tropics into the high and mid latitudes of the cooled hemisphere (Chiang and Bitz 2005; Cheng et al. 2007). Thus, quoting Chiang and Bitz 2005: “From an energetics viewpoint, the ITCZ displacement can be interpreted as the required response of the model Hadley circulation in transporting atmospheric energy northwards to compensate for the loss of energy in the high latitudes.” The ITCZ shift appears to be a robust response to high latitude cooling, independent of the background climate (Cheng et al. 2007), although the sensitivity of the atmospheric transport response varies between the interglacial and glacial conditions (Cvijanovic et al. 2011).

The second perspective, namely the energy flux perspective, posits that the tropical ITCZ shifts are tied to the cross-equatorial atmospheric energy transport changes induced by the high latitude cooling. Kang et al. (2008, 2009) viewed the energy flux changes as key to interpreting the tropical ITCZ response to high latitude cooling, and developed a framework to explicitly link them to the tropical precipitation changes. They investigated the global energy budget response to extratropical thermal forcing in a set of idealized aquaplanet experiments. Forcing applied in these two studies resembles the “bipolar seesaw” pattern, with cooling in one and warming in the other hemisphere, with the global average forcing equal to zero. They find a close connection between the ITCZ shifts and cross-hemispheric energy transports, and interpret this link as follows. Their applied forcing implies an anomalous cross-hemispheric oceanic heat transport, which is compensated to a degree by the anomalous cross-hemispheric atmospheric heat transport resulting from the atmospheric response to the imposed forcing. Given the level of compensation and the ambient tropical gross moist stability, the change in the cross-equatorial mass flux (and hence tropical precipitation response) can be predicted. They argue for a dynamically-determined level of compensation (presumably through altered eddy fluxes) of around 25 %, based on aquaplanet-model calculations with no cloud and water vapor feedbacks. The addition of cloud and water vapor feedbacks, however, brings up the level of compensation up dramatically, up to around 115 % in the atmospheric general circulation model used in their study. These studies thus also illustrate the sensitivity of the tropical response to model convection scheme and cloud feedbacks. In overall, the energy flux perspective, as described in Kang et al. (2008, 2009) studies, appears to represent a useful diagnostics for determining the tropical precipitation response.

In a follow-up study, Kang and Held (2011) investigated to importance of WES feedback in tropical precipitation shifts to extratropical thermal forcing and demonstrated that the ITCZ shifts are also possible without the WES feedback. Moreover, in a series of idealized experiments with WES feedback suppressed, they have also shown that the tropical precipitation response remains similar, even though the simulated tropical surface temperature gradient changes are smaller. According to this study, the atmospheric energy budget and the moist static stability determine the precipitation response in the model rather than tropical SSTs. This interpretation had contributed to further extension of the energy flux perspective by suggesting an alternative mechanism for the tropical precipitation shifts. In their view, the SST response should be seen as a result of the surface energy budget and not as a driver of the precipitation response (Kang and Held 2011). This statement is, however, in direct contrast with the SST view according to which the SST changes drive the ITCZ shifts.

Reconciling the SST and energy flux perspective is certainly not trivial, but understanding the weaknesses and strengths of each perspective would advance our knowledge of the high to low latitude teleconnections. The main disagreement between the two lies in the proposed mechanisms of the ITCZ shifts (SSTs versus the atmospheric energy flux changes). However, it is important to note that there is a significant overlap between the two views, making them also complementary in many aspects. In that sense, they both agree the ITCZ shifts to the warmer hemisphere in order to allow for the increased energy transport to the cooled hemisphere (Chiang and Bitz 2005; Kang and Held 2011). Thus, in both perspectives the ITCZ shift is associated with the change in atmospheric heat transport. Also, both perspectives suggest that the meridional SST gradients play a role in the tropical precipitation adjustment. However, it remains unclear how this later point concurs with another conclusion from Kang and Held (2011) study stating that the different SST patterns can cause the same precipitation shifts, as the later would imply that the precipitation shifts are possible without the SST changes. One of the aims of this study is thus, to provide a conclusive answer to the question whether the ITCZ shifts are possible without the tropical SST changes.

This study examines the response to high-latitude North Atlantic cooling in an atmospheric general circulation model (AGCM) coupled to a slab ocean model. In comparison to the previous studies like Kang et al. (2009) and Kang and Held (2011), our simulations are not performed in the aquaplanet mode, but use realistic surface boundary conditions. Another point is that the imposed cooling is applied only in the high-latitude North Atlantic and it is not ‘compensated’ in the southern hemisphere. This scenario is based on abrupt climate change events during the last glacial that are thought to be driven by North Atlantic cooling. As the atmospheric response timescales are much shorter than oceanic, our approach reflects the global atmospheric adjustment after the high latitude cooling has been imposed. Using this semi-realistic test case, we examine various aspects of the tropical response to the North Atlantic high-latitude cooling in the context of the SST and energy flux perspectives. In particular, we examine (1) the origins of the top-of-the-atmosphere (TOA) energy budget changes by evaluating the radiative feedback effects; and (2) we further investigate the relative roles of the atmospheric energy budget and SST changes in the ITCZ shifts, following the arguments by Kang and Held (2011).

The global TOA energy budget changes to high latitude cooling are examined by considering the extratropical, northern and southern tropical energy budget anomalies and decomposing their corresponding TOA flux changes into individual feedback effects. In particular, the radiative kernel technique of Soden et al. (2008) is used to quantify individual shortwave and longwave feedbacks arising from the temperature, cloud, water vapor and albedo changes. Shell et al. (2008) and Zelinka and Hartmann (2011) previously used the radiative kernel method to assess the role of different feedbacks in global warming experiments. Another study by Yoshimori and Broccoli (2009) used a ‘partial radiative perturbation’ method to assess the radiative feedbacks in AGCM-slab ocean simulations with reduced solar irradiance in the northern extratropics. They found the energy budget changes to be chiefly influenced by the water vapor and lapse rate feedbacks, with cloud feedbacks playing the role in cooling of the northern tropics and midlatitudes. Our study, on the other hand, provides the quantitative radiative feedback analysis for simulations with the imposed high latitude North Atlantic surface cooling. We focus on the atmospheric reorganization that provides the energy compensation for the high latitude energy loss, and identify the main radiative feedbacks and the locations responsible for the energy gain.

The relative roles of SST versus the atmospheric energy flux changes in the ITCZ shift are investigated in a series of idealized experiments that allow us to separate the influence of the remote (high latitude) and local SST forcing on tropical circulations. Our results emphasize the constraints that tropical SST changes impose on the ITCZ: while the energy flux perspective provides a useful diagnostic for predicting the magnitude of the tropical rainfall change, the actual mechanism of the shift requires the SST perspective. The manuscript is organized as follows: in Sect. 2 we introduce the model and the experiments. Results are presented in Sect. 3 by first considering the global energy budget changes in the slab ocean experiments. In Sect. 4, we describe the hypothesis on the importance of tropical SSTs in the final energy compensation. This hypothesis is then tested in series of idealized experiments with fixed tropical SSTs in Sects. 4.1 and 4.2, while in Sect. 4.3 the results of transient experiments are briefly discussed. Section 5 summarizes the conclusions and discusses the possible implications.

2 Model setup and simulation design

We use the NCAR’s Community Atmosphere Model version 3 (CAM3) in the slab ocean mode, with T42 resolution and 26 vertical levels. The physics and dynamics of the CAM3 model are described in Collins et al. (2004, 2006). The model includes the Community Land Model (CLM) version 3.0 (Bonan et al. 2002; Oleson et al. 2004). Spatially varying annual mixed layer depths for the slab ocean are derived from the Levitus data (1982), and a monthly-varying ocean heat flux convergence (aka the Q-flux) is applied to keep the simulated SST close to present-day. Present day land surface conditions, orbital and greenhouse forcing are applied. High-latitude North Atlantic cooling is imposed by altering the Q-flux field in the manner as shown in Fig. 1.
Fig. 1

Geometry of the Q-flux anomaly applied over the North Atlantic sector

We undertake two sets of simulations, one where the slab ocean is applied globally, and the other set where there are mixed surface ocean boundary conditions—fixed SSTs in the tropics, and slab ocean elsewhere. We start by running the control (hereafter labeled ‘CTRL’) and the North Atlantic cooling (‘SOM1’) cases in the CAM3 coupled to the slab ocean (a summary of all the simulations done is listed in Table 1). For the SOM1 simulation, the magnitude of maximal forcing equals 15 W/m2 (cooling) and it is applied as shown in Fig. 1. In these slab ocean simulations, the model was run for 40 years with the first 10 years of output discarded as spin up.
Table 1

Experiment summary


Fixed tropical SSTs

NA qflux forcing






Yes, 15 W/m2


Yes, as in CTRL



Yes, as in CTRL

Yes, 15 W/m2


Yes, as in SOM1



Yes, as in SOM1

Yes, 15 W/m2

To assess the role that tropical SST changes played in determining the ITCZ response, we take the SSTs generated by these two simulations, and impose them in the CAM3 model simulations as a tropical fixed SST boundary condition (see next paragraph for specifics). The fixed SSTs are applied over the global tropics from 26.5°S to 26.5°N. South and north of 32.1°S and 32.1°N respectively, a slab ocean is applied. In the transition regions (26.5°–32.1° in either hemisphere), the SST values are nudged between the fixed and slab ocean (calculated) values. More precisely, the surface temperature values over the ocean surface points in the transition regions are obtained as a weighted average of the fixed and calculated SST values with the weights of the calculated SSTs increasing linearly from 0 at 26.5° till 1 at 32.1° and the opposite for the fixed SST weights.

In total, there are four idealized experiments with the SSTs fixed in the tropics (and slab ocean elsewhere): (1) FCTRL—unperturbed tropical SSTs corresponding to the ones from the control slab ocean experiment, and no North Atlantic cooling applied to the slab ocean; (2) HL—fixed climatological tropical SSTs as in FCTRL, plus the North Atlantic high latitude cooling; (3) TROP—fixed tropical SSTs corresponding to those simulated by the slab ocean experiment with North Atlantic cooling, but with no North Atlantic high latitude cooling applied to the slab ocean and (4) HL + TROP—fixed tropical SSTs as in TROP, plus the North Atlantic high latitude cooling applied to the slab ocean. In these idealized experiments, the model was run for 60 years, with output of the first ten discarded as spin up.

In addition to this, we have performed a set of transient simulations that have enabled to follow the adjustment period in more detail and test the conclusions derived from the idealized experiments. Presented climatologies for the transient simulations represent the mean of 20 ensemble members, obtained by branching from the initial control run each 5 years and applying the North Atlantic cooling.

3 High latitude and tropical response to North Atlantic cooling

3.1 General characteristics of the equilibrium response

Annual surface temperature anomalies due to imposed high latitude cooling are shown in the Fig. 2 (upper panel) for SOM1 experiment. Similar to Chiang and Bitz (2005) and Mahajan et al. (2011) we find that the high latitude cooling has spread into the northern tropics, forming a dipole with colder northern and warmer southern tropics. This is especially pronounced over the Pacific sector. Maximal (annual) cooling is about 8 °C in the high latitudes and 0.5 °C in the northern tropics while the southern tropics became warmer by about 0.5 °C.
Fig. 2

Annual mean surface temperature and precipitation anomalies in the SOM1 experiment

Precipitation anomalies (Fig. 2, lower panel) show a southward shift of the maximal precipitation zone over all ocean basins, with ~2 mm/day increase over the southern tropics and a similar decrease over the northern tropics, indicating a southward shift in the ITCZ. Precipitation also decreases by ~1.5 mm/day in the North Atlantic, where the cooling was imposed. The mean meridional stream function anomaly (Fig. 3, left panel) indicates a southward displacement of the uplift region, leading to a strengthening of the northern Hadley cell and weakening of the southern cell. Surface wind anomalies (Fig. 3, right panel) reveal an increase in the north-easterly trades and a decrease in the south-easterly winds.
Fig. 3

Left plot: Annual meridional stream function anomaly (kg/s), contour lines are shown every 1011 kg/s and right plot: annual surface wind speed anomalies (m/s) for the SOM1 experiment

We now examine the energy budget and poleward energy transport changes (Fig. 4). The surface energy budget (black dashed line) is altered essentially only over the latitudes of imposed cooling. The top-of-the-atmosphere (TOA) energy budget (solid black line), on the other hand, shows change across all the latitudes, with a maximal anomaly of 0.06 GW/m in the southern tropics. The northward atmospheric heat transport anomalies (Fig. 9) show increased transport from the southern tropics to the northern mid and high latitudes. The additional energy deposited in the northern mid and high latitudes originates from increased TOA energy flux in the southern tropics. If we increase the magnitude of the high latitude forcing, we find the same qualitative response in rainfall and energy fluxes, but with increased magnitude (not shown).
Fig. 4

Top-of-the-atmosphere (solid) and surface (dashed) annual zonal energy budget anomalies (normalized by the length of the latitude circle)

Similar features of the tropical response due to the North Atlantic cooling were found in study by Chiang and Bitz (2005), in simulations where increased land ice or sea ice cover were applied in the Northern hemisphere. It has also been noted previously, in this and other studies (Chiang et al. 2003; Mahajan et al. 2011) that the southern tropics tend to be the location where increased TOA energy flux arises providing the additional energy flux to the northern high latitudes. Our aim is to broaden these conclusions by addressing the exact feedback mechanisms that act to provide the TOA energy surplus locally and remotely and offer an explanation why the southern tropics are favorable area instead of, for example, northern tropics. We continue our analysis by investigating how the high latitude energy demand propagates to the tropics and what would be the requirement for the energy compensation to occur somewhere else.

3.2 Global energy budget and radiative feedbacks response

3.2.1 Radiative kernel technique

We apply the radiative kernel technique developed by Soden et al. (2008) and implemented for the CAM3 by Shell et al. (2008) which allows us to decompose the TOA changes into various climate feedbacks. The technique separates the individual climate feedbacks into two parts: the radiative kernel term \( \frac{\partial R}{\partial X} \) and the climate response of the feedback variable \( \frac{dX}{{dT_{s} }} \). Here \( \partial R \) refers to the total change at the TOA, X is the considered climate variable and dTs is the change in the global mean surface temperature (in a steady state). Defined as such, the radiative kernel quantifies the change in TOA flux due to small perturbations in the corresponding climate variable. Given the climate response (from the model output) we can combine the two and calculate the individual feedback parameters, while the total feedback parameter γ is considered a sum of the individual feedbacks and a small residual Re:
$$ \gamma = \sum\limits_{i = 1}^{n} {\frac{\partial R}{{\partial X_{i} }}\frac{{dX_{i} }}{{dT_{s} }}} + \text{Re} $$
The total change in the top of the atmosphere energy budget can then be approximated as:
$$ \Updelta R = \gamma \Updelta T_{S} = (\gamma_{T} + \gamma_{q} + \gamma_{alb} + \gamma_{c} )\Updelta T_{s} $$
where γT, γq, γalb and γc are the temperature, water vapor, surface albedo and cloud radiative feedbacks. Shortwave and longwave radiative kernels are provided for most of the climate variables except for the clouds. Due to the non-linearities associated with the cloud feedbacks, the later are calculated by applying the corrections to calculated cloud radiative forcing (CRF). For details on the feedback calculations, we refer the reader to Soden et al. (2008) and Shell et al. (2008).
Given that the radiative kernels are calculated around a predetermined present-day base state climate that may differ in details from our simulations and that their application was mainly tested in the CO2 doubling experiments, we first evaluate the technique by considering only the clear sky feedbacks (Fig. 5a, b). Figure 5a shows that there is an excellent agreement between the TOA longwave energy flux changes (solid red line) and the sum of longwave feedback effects (dashed red line) comprising of those for temperature, surface temperature and water vapor. Shortwave feedbacks (dashed red line, Fig. 5b) do not show as good agreement with the total TOA shortwave energy budget changes (solid red line, Fig. 5b). This difference mainly occurs in the high latitudes, suggesting that the cause of disagreement is the shortwave surface albedo feedback by sea ice. Shell et al. (2008) discusses the uncertainties related to the radiative kernel technique and its limitations in surface albedo feedback calculations. They noted that this parameter tends to behave more nonlinearly compared to the other feedbacks, with the effect being especially pronounced at the edges of sea ice cover.
Fig. 5

Clear-sky (a, b) and all-sky (c, d) longwave (a, c) and shortwave (b, d) feedback effects and TOA energy budget anomalies in SOM1 experiment. Black lines: shortwave feedback effects [water vapor (solid), albedo (dot dashed) and cloud (dotted)] and longwave feedback effects [temperature—surface and air (dashed), water vapor (solid) and cloud (dotted)]. Red lines: sum of corresponding feedback effects (dashed) and the respective TOA energy budget anomalies (solid)

All-sky feedbacks, shown in Fig. 5c, d, behave in similar manner with a very good agreement between TOA longwave changes and the sum of the longwave feedback effects but with lesser, but still reasonable, agreement between the shortwave feedback effects and TOA shortwave changes in the high latitudes. For this reason, we are cautious in interpreting high and midlatitude shortwave feedbacks. Despite of that, the comparison of the zonal TOA energy budget anomalies and the overall feedback effects in Fig. 5 suggest that the disagreement between the two is much smaller than the actual anomalies and therefore is not likely to significantly alter our findings.

3.2.2 Energy balance decomposition

The principal high latitude effect originates from the temperature longwave feedback (Fig. 5c, d). It consists of both the surface and air temperature feedbacks, and it is a negative feedback as it acts to damp the initial cooling. In the tropics, the majority of TOA energy budget changes are due to the longwave and shortwave cloud and longwave water vapor feedbacks. Longwave cloud feedbacks act to warm the southern and cool the northern tropics, while shortwave cloud feedbacks have an opposite effect.

In order to summarize the essential features of the energy flux changes and the role of radiative feedbacks, we simplify the considered regions into three distinct latitude bands: 90°N till 30°N (northern extratropics), 30°N till 10°N (northern tropics) and 10°N till 40°S (southern tropics). These bands were chosen based on an assessment of regions that exhibit similar behavior, specifically using the sign of the zonal TOA net energy budget anomaly. The analysis is shown in Fig. 6a where all the fluxes are being defined positive into the box.
Fig. 6

Schematic box energy budget—TOA and surface net energy budget anomalies are given at the top and bottom of each box (respectively), anomalous atmospheric heat transports are shown at the borders between the boxes and the main feedback effects are listed within the each box. All values are normalized by the surface energy budget anomaly from box 1, fluxes are defined positive into the box. a SOM1 experiment, b HL experiment. Legend: temp—temperature (including surface temperature), wv_lw—water vapor longwave, wv_sw—water vapor shortwave, crf_sw—cloud shortwave, crf_lw—cloud longwave feedback effects; toanet—TOA net energy flux, surfnet—surface net energy flux

Applied North Atlantic cooling causes a surface energy imbalance of −0.191 PW over the northern extratropical box (90°N–30°N). We use the anomalous surface energy budget from this box to normalize the energy flux and transport changes in this and other regions, making the surface loss in the northern extratropical box equal −1 (−100 %). It is important to mention that this surface energy loss encompasses both the applied cooling and the surface feedbacks. Figure 6a further shows that 36 % of this initial loss is compensated by the TOA energy budget changes over the same region (0.069 PW). The TOA response results from a complex mix of negative and positive feedbacks. The longwave temperature feedback has the strongest effect, and with 0.256 PW (134 %) is more than the amount lost to the surface. Countering this are the positive feedbacks that act to cool the area, with the biggest contributions coming from water vapor longwave −33 % (−0.064 PW), cloud shortwave −26 % (−0.049 PW) and albedo −29 % (−0.036 PW) feedback effects. The overall energy flux between TOA and the surface in the northern extratropical box is negative, and to maintain energy balance the atmosphere needs to import energy from the lower latitudes at a rate of 0.121 PW (64 %) across 30°N. In summary, because of the large cancellation between the negative and positive radiative feedbacks, the majority of the surface cooling has to be balanced by northward atmospheric transport.

We now consider the northern tropical box (10°N–30°N). The surface flux changes are essentially zero (less than a percent), as it has to be because of the slab ocean constraint. The TOA net radiative flux shows a small loss of 3 %, which is in agreement with the coupled atmosphere–ocean experiments by Cheng et al. (2007) who also noted small negative TOA anomalies over the northern tropics. Main changes in the TOA energy budget in this box originate from the cloud feedbacks due to the southward ITCZ shift and longwave water vapor feedbacks due to the cooling and drying of the northern tropics. Negative TOA flux anomaly arising from the cloud and water vapor longwave feedback effects is larger than the positive anomaly due to the shortwave cloud and temperature feedbacks. Overall, the TOA flux changes are small, but negative, and the northward energy transport of 0.121 PW out of the northern tropics is compensated for by somewhat larger atmospheric transport from the southern tropics of 0.128 PW (67 %).

In the southern tropical box (40°S–10°N), the northward flux across its southern boundary is essentially zero, meaning that the energy required to supply the energy flux out of the northern edge has to come from TOA flux changes within the box. This positive TOA energy budget anomalies result from the longwave feedback effects arising from clouds 40 % (0.076 PW), water vapor 19 % (0.037 PW) and temperature 14 % (0.027 PW). The total effect of the shortwave cloud radiative feedbacks over this area is small and negative (−6 %).

We summarize the global energy budget response as following: the initial northern high latitude surface cooling is first compensated by the local TOA changes (about one-third) while the rest of the energy is transported from the lower latitudes. The northern tropics do not provide required energy to close the energy budget, this is because despite the fact that there are sizable changes in TOA longwave and shortwave fluxes, they essentially cancel each other out. Rather, it is the southern tropics that provide the energy required, which is transmitted by atmospheric energy transports across the northern tropics to mid and high latitudes.

Tropical top-of-the-atmosphere energy flux changes result from the three main feedback effects, originating from clouds, water vapor and temperature. Temperature and water vapor feedbacks have opposing effects: over the colder and consequently drier areas, temperature feedback provides positive and water vapor negative TOA energy flux changes. In the high northern latitudes, the temperature feedback is stronger, so the overall sign of the TOA anomaly is determined by the temperature feedback. In the tropics on the other hand, the strongest feedback effect comes from the clouds due to the ITCZ shift and it exerts two opposing effects. The longwave cloud feedback induces positive TOA energy budget anomalies in the southern hemisphere and negative in the northern hemisphere, while the shortwave cloud feedback has the opposite effect, but smaller, making the overall TOA anomalies in the tropics follow the sign of the longwave cloud feedback effect. Overall, the colder and drier northern tropics are not capable of providing significant positive TOA energy flux changes (as shown earlier they are actually negative) due to the cancelation between longwave and shortwave feedback effects. In contrast, warmer and wetter southern tropics with the southward-displaced ITCZ, represent the perfect location to gain the energy required due to the high latitude energy loss.

Our results differ from Kang et al. (2009) in sense that shortwave cloud feedback dominates over the longwave cloud feedback in the tropics in their aquaplanet simulations. Consequently, the overall cloud feedback effect induces positive TOA flux anomalies in the northern tropics and negative in the southern tropics, an outcome attributed to increased low cloud cover over the northern tropics. In the next chapter we provide a possible explanations for this discrepancy, and clarify the relationship between the tropical SST anomalies and the precipitation shifts.

4 Proposed interpretation

Despite the fact that the southward ITCZ shift is causing a small amount of energy loss in the northern tropics due to the cloud feedbacks, it appears advantageous for the atmosphere to move the ITCZ south, because of the large energy gain accomplished by the convection reorganization there. The question we ask here is: why the southern tropics and not somewhere else? Inspection of the tropical TOA anomalies (Fig. 7, top panel—contours) shows, more accurately, that the areas of maximal (positive) TOA budget changes are located in the east tropical Pacific, encompassing the Pacific cold tongue area.
Fig. 7

Upper plot: TOA flux climatology (colors) and its anomalies (contours) in SOM1 experiment, lower plot: tropical TOA flux anomalies (colors) overlayed with high cloud cover anomalies in the SOM1 experiment. Contours: full lines—positive anomalies, dashed—negative anomalies, contour interval equals 0.02 for the high cloud anomalies and 2 W/m2 for the TOA energy budget anomalies, zero contour ommited

Interestingly, they resemble the areas of decreased OLR during warm El Nino-Southern Oscillation (ENSO) events (Su and Neelin 2002) leading us to speculate that the tropics absorb additional energy through a convective reorganization that resembles that of an El Nino event. Deep tropical convection regions have more TOA net energy flux towards the earth than comparable regions with only shallow trade cumulus. While high anvil clouds associated with deep convection may reflect more shortwave back into space, this is more than compensated for by the reduced outgoing longwave radiation because of the increased cloud height. Indeed, Fig. 7 (bottom panel) confirms that the areas of strong positive tropical TOA anomalies (colors) are coincident with the increased high cloud cover (contours).

Tropical convection is known to be “tippy” as there is a threshold for deep convection, such that small changes in temperature (and consequently, moisture) at the margins of deep convective regions (so called ‘convective margins’) can initiate (or halt) deep convection. Locations with large zonal and meridional temperature gradients neighboring deep convection regions are such convective margins. During El Nino, warmer SSTs over the cold tongue leads to SST regions neighboring the Pacific warm pool to become sufficiently warm to exceed the threshold for deep convection.

In our case, however, convection is initiated over the marginal regions as a result of both the warmer SSTs in the tropical east Pacific region and colder mid-troposphere above it that lowers the threshold for convection. As previously noted by Chiang and Bitz (2005), weakened easterlies over the south-east tropical Pacific give rise to the initial SST warming while the SST dipole is maintained due to the higher water vapor content in the warmed hemisphere. Although the observed warming is smaller than during the El Nino events, combined with lowered convection threshold it is sufficient in triggering the convection over the marginal regions. This additional deep convection provides the necessary increase in the TOA energy flux that makes up for the high latitude energy loss. Observed changes in global TOA absorption (Wong et al. 2009) shows that the convective reorganizations related to ENSO events are indeed capable of affecting the global energy budget.

In comparison with the experiments by Kang et al. (2009), there are several differences to our setup that may have attributed to different conclusions, mainly the use aquaplanet model with the zonally uniform SSTs and the differences in cloud parameterizations. Our experiments show that locations with zonal asymmetries can similarly respond to high latitude cooling, whereas in aquaplanet simulations there are only meridional asymmetries. In addition, the model applied in our study responds by increasing the high cloud cover, in contrast to the predominant low cloud changes in Kang et al. (2009). Since our proposed interpretation favors the SST perspective, we further focus on the significance of tropical SST changes in ITCZ shifts in series of idealized experiments.

4.1 Tropical SST anomalies as fundamental to ITCZ shifts

In order to investigate the role of tropical SSTs in ITCZ shifts, we perform idealized simulations in which the tropical SSTs are fixed to specified values, while keeping the slab ocean lower boundary condition in the extratropics (see Sect. 2 for details). In the control experiment (FCTRL), the tropical SSTs are fixed to climatology, and no high latitude North Atlantic cooling is applied. In the forced experiment (HL), high latitude North Atlantic cooling is applied in the same way as in the full slab ocean experiment, but tropical SSTs are not allowed to respond to these changes.

Surface temperature anomalies from HL experiment are shown in the upper panel of Fig. 8. The cooling is limited to northern high and mid latitudes except for the small progression of cold anomalies into the tropics over land. Compared to the full slab ocean case, there is less cooling over the northeastern Asia while the cooling pattern over the northwest America is reversed and the south-east remains warm.
Fig. 8

Annual mean surface temperature (upper plot) and precipitation (lower plot) anomalies in idealized experiment HL (tropical SSTs fixed to unperturbed values)

Annual precipitation anomalies are shown in the lower panel of Fig. 8. The high latitude precipitation response features decreased values over the North Atlantic (where the cooling was imposed), similar to SOM1 case. In contrast, tropical precipitation changes are essentially not present, and show no southward ITCZ shift. This confirms that SST changes are necessary for the ITCZ shifts.

Northward heat transport anomalies, shown in Fig. 9, reveal that the energy is transported from the northern tropics into the high latitudes, making the northern tropics the source of the additional energy. In the cases with fixed tropical SST, increased surface fluxes from the neighboring mid-latitude and tropical ocean area are at least as important as the TOA flux changes, when considering the energy absorbed in the column.
Fig. 9

Northward atmospheric heat transport anomalies (PW): first column—SOM1 and the idealized experiments HL, TROP and HL + TROP, second column—transient experiments (years 1, 2, 3 and 7)

Figure 6b, showing the box model, summarizes the way in which the energy flux changes come about in the HL experiment. In the northern extratropical box, there is a loss of −0.15 PW to the surface, due to the imposed cooling. As in the previous case, this will be used as a normalization factor, so that the total loss of energy to the surface in this box equals −1. Local TOA energy budget changes balance 25 % of this initial energy loss to the surface. As in the full slab ocean case, temperature acts as the strong negative feedback, while the other feedbacks act to increase the initial cooling. The energy budget in the northern extratropical box is balanced by a northward energy transport of 0.11 PW (75 %) from the northern tropics. Most of the energy required for this anomalous northward heat transport comes from the northern tropical box itself, and mainly from the surface (59 %), while the most of the remainder is supplied from the southern tropics (10 %). A small amount is also acquired through positive TOA anomalies due to longwave feedbacks (5 %). The southern tropics, in turn, gain most of the energy through northward transport across 40°S.

The fixed tropical SSTs (set at climatology) provide an infinite heat source. As the cold air from the high and mid latitudes advects southwards into the northern tropics, the air-sea temperature difference over the northern tropics increase, resulting in increased surface fluxes into the atmosphere, in this case providing the bulk of the energy required by the northern extratropics. The increased surface fluxes arise largely from increased evaporation. In this idealized experiment, the crucial difference from the full slab-ocean setup is that surface flux changes are providing the bulk of the energy to supply the northern extratropics; whereas in the full slab ocean experiment this energy can only come from the TOA flux changes (chiefly due to the cloud radiative feedback effects over the east tropical Pacific).

The main conclusion to be drawn from this idealized experiment is that tropical SST anomalies are crucial in order for the ITCZ to shift. From the energetics viewpoint, the first place where the atmosphere tries to obtain the increased energy supply from is the northern tropics, specifically from the surface fluxes. In the fixed SST experiment, this supply can be maintained because it is an infinite source. In the slab ocean (and presumably the real world), this source is finite, and the subsequent northern tropical SST cooling leads to the southward ITCZ shift and relocation of tropical convection, resulting in the southern tropical TOA energy gain. In both the fixed-SST and slab ocean cases, TOA fluxes in the northern tropics do not provide the needed energy. In the fixed SST scenario, the northern tropical climate does not change much and so the temperature, water vapor and cloud feedbacks are much smaller. As such, the northern tropical TOA energy budget changes are again not the main contributor to the energy gain.

4.2 Locally driven ITCZ shifts

The previous experiment showed that the tropical SSTs are required to change in order for the ITCZ to shift. We now ask whether the tropical SST changes are sufficient for the ITCZ to shift, regardless of whether high latitude North Atlantic cooling is imposed. To examine this, we expand our analysis with two new experiments. In both experiments, tropical SSTs are fixed and set to values corresponding to those in the full-slab ocean simulation with imposed cooling. However, in the extratropical slab ocean, one has no high-latitude North Atlantic cooling (hereafter the TROP experiment), while the other does (hereafter the HL + TROP experiment). We compare these simulations to the same control run as before (FCTRL).

The annual mean temperature and precipitation anomalies in the two experiments are shown in Fig. 10. Temperature changes in HL + TROP case resemble the full slab ocean case, while TROP case exhibits cooling over the North Atlantic and warming over the northern Euro-Asia and northern Pacific area. In both simulations, however, the tropical precipitation anomalies closely resemble that for the full-slab ocean (SOM1) case, both in terms of structure and magnitude. We compared the annual and zonal mean precipitation anomalies for cases TROP, HL + TROP and SOM1 in Fig. 11a. Compared to the fixed SST cases, SOM1 has slightly smaller precipitation anomalies around the equator, but more pronounced second precipitation peak and a wider zone of positive precipitation anomalies. TROP and HL + TROP show very similar equatorial precipitation responses, suggesting the same ITCZ shifts in the two simulations.
Fig. 10

Annual mean surface temperature (upper plots) and precipitation (lower plots) anomalies in TROP and HL + TROP experiments
Fig. 11

Annual mean zonal precipitation (a) and TOA flux (b) anomalies in SOM1, TROP and TROP + HL experiments

The northward atmospheric heat transport anomalies in the HL + TROP case (Fig. 9) resemble that for the full-slab ocean (SOM1) case: an atmospheric energy gain from the southern tropics is transported to the northern extratropics, with comparable magnitudes of the poleward flux anomalies. Zonal TOA energy budget changes for HL + TROP are also in very good agreement with the ones from the SOM1 experiment (Fig. 11b). This is not surprising as the climate conditions are essentially the same between the two runs.

The TROP simulation shows a significantly different character in the northward energy transport change—while atmospheric energy is gained in the southern tropics as in the HL + TROP case, this energy is essentially transported to, and mainly ‘lost’ in, the northern tropics. Relatively small anomalous northward heat transport exists from the tropics to the high and mid latitudes and it is a consequence of the overall energy surplus in the area 10°N–30°N. On the other hand, zonal profiles of the tropical TOA energy budget anomalies (Fig. 11b) show almost identical changes in TROP and HL + TROP idealized cases. The two not only have the same tropical SST patterns, but also very similar precipitation and TOA energy budget response in the tropics (as seen from Fig. 11a, b). Thus, TROP simulation suggests that the tropical SST changes alone are sufficient to shift the ITCZ southwards and provide the same energy surplus, independently of the actual high latitude energy demand.

In the deep tropics, the same SST changes are associated with practically the same precipitation responses (Fig. 11a) in TROP and HL + TROP experiments, in agreement with the SST perspective. However, the atmospheric heat transport responses (Fig. 9) averaged between 10°S and 10°N are also similar in these two experiments and the energy flux perspective is satisfied. Thus, our results confirm the usefulness of the energy flux perspective in determining the tropical precipitation response.

Regarding the mechanism of the precipitation shift, Figs. 9 and 11a suggest that there is a much closer agreement between the precipitation changes than between the atmospheric heat transport anomalies over the deep tropics in TROP and HL + TROP. Moreover, the two experiments show very similar TOA flux responses, implying that the different tropical atmospheric heat transport responses are due to the different surface flux responses. The same (imposed) SST anomalies are thus associated with the same precipitation changes and only similar atmospheric heat transport and surface heat flux responses. In conclusion, in contrast to the results of Kang and Held (2011), in these experiments, the tropical SSTs changes appear to be a more suitable candidate for driving the tropical precipitation changes compared to the surface fluxes.

4.3 Establishment of the anomalous northward heat transport

We now analyze the transient adjustment to the high-latitude North Atlantic cooling to support our interpretation of how the tropical energy gain and northward atmospheric heat transport changes come about. This realistic setup can provide more certainty in the results obtained in our idealized experiments and possibly bring an insight about the mechanism of the shift by reveling whether the changes in SSTs or precipitation come first.

The first 3 years (establishment) and the year 7 (developed response) in the transient adjustment of the anomalous northward atmospheric heat transport to high latitude cooling are shown in Fig. 9. During the first two years, most of the additional energy transported northward comes from the mid-latitudes and northern tropics. A noticeable anomalous northward mid-latitude transport establishes from year 2 (Fig. 9b). This comes about due to the combined effect of the temperature and cloud feedbacks that cause positive midlatitude TOA budget anomalies as well as the fully developed high latitude surface forcing (not shown). A small energy surplus arises in the southern tropics too, as a result of warming due to the longwave cloud feedbacks. During the year 3, positive TOA energy flux anomalies in the southern tropics increase as the cloud and temperature feedbacks develop.

Figure 12, showing the tropical surface temperature and precipitation changes during the adjustment period (years 1, 2, 3 and 7), reveals a development of the northern hemispheric cooling and southern tropical warming, and a southward precipitation shift (especially pronounced over the Indian Ocean during the year 3). In the eastern Pacific, this signal is lost during the year 4 and recovered again in year 5, when the major energy gain comes from the southern tropics (not shown). In the following years, stronger surface temperature dipole and more striking southward precipitation shifts develop and the southern tropics remain the area of significant energy gain. The structure of dominant feedbacks starts to resemble the equilibrium response from year 5, with the positive TOA anomalies in the southern tropics due to the cloud, water vapor and temperature feedbacks (not shown).
Fig. 12

Left column: Tropical surface temperature anomalies (K) and right column: tropical precipitation anomalies (mm/day) in the transient period (years 1, 2, 3 and 7)

With regards to the role that SST plays in the precipitation response, the transient experiments show essentially simultaneous changes in the surface temperatures and precipitation. As such, it is not possible to infer which comes first (precipitation or SST change). Nevertheless, these experiments are not in disagreement with our interpretation of the idealized experiments that the tropical SST anomalies drive the precipitation shifts (Sect. 4.2), as there is no indication of the precipitation shifts without the tropical SST anomalies present.

5 Summary and discussion

In this study, we examined in detail the global energy budget changes to high latitude North Atlantic cooling and the southward ITCZ shift in an atmospheric general circulation model coupled to a slab ocean. Our energy budget approach follows from previous studies that suggest the utility of this approach, in particular Kang et al. (2009). First, we have evaluated the energy budget changes in extratropics, northern and southern tropics, and quantified the individual feedbacks responsible for them using a radiative kernel technique. Second, we examined the interplay between energy fluxes and SSTs in determining the ITCZ response to high latitude North Atlantic cooling, specifically contrasting the two existing perspectives—the role of SSTs, and the role of energy fluxes in interpreting our simulations.

We found that 35 % of the initial high latitude North Atlantic surface cooling is compensated by TOA changes in the northern extratropics, and mainly due to the longwave temperature feedback, while the rest of the energy required to balance the energy flux is provided from the southern tropics, mainly due to the longwave cloud feedback. The northern tropics essentially did not contribute to the net energy fluxes, because of a cancellation between longwave and shortwave feedbacks.

We also carried out idealized simulations, by fixing SSTs in the tropics while keeping the slab ocean boundary condition in the extratropics, in order to examine the role of SST anomalies in the tropical ITCZ response. Our results suggest that the tropical SST changes directly control the ITCZ response, whereas there is no direct link to the extratropical energy demand. Only as the cooling anomalies reach the tropics (due to progression of the surface temperature anomalies), does the convection reorganize; this reorganization then leads to TOA energy flux changes from cloud, water vapor and temperature feedbacks. Our result thus suggests that the SST anomalies propagating into the tropics are both necessary and sufficient for the ITCZ to shift, in accordance with the ‘SST perspective’ on ITCZ shifts.

The Kang et al. (2008) study reports that the overall cloud effect is negative in the southern tropics and positive in the northern tropics with the shortwave component being dominant over the longwave. This result differs from our experiments, which show that longwave cloud feedback is at least as important as the shortwave cloud component (Fig. 12a). Furthermore, in our experiments the longwave cloud feedbacks together with the temperature and water vapor feedbacks determine the sign of TOA anomalies over most of the tropics. In general, total (shortwave and longwave) cloud radiative effect is positive over the southern tropics. This discrepancy is likely to be caused by the differences in the convective and cloud parameterizations between our model and the ones used in Kang et al. (2009) study.

The applied extratropical forcing in the Kang et al. (2008) aquaplanet study was cooling in one hemisphere, and an equal but opposite warming in the other hemisphere. On the other hand, our simulation (that uses more realistic boundary conditions) has an imposed high-latitude North Atlantic cooling that is not ‘compensated’ in the southern hemisphere. The North Atlantic cooling scenario is motivated by the scenario of AMOC slowdown. Given that the time scale of the atmospheric response is much shorter than for the ocean, our particular AGCM-slab ocean experiments is relevant to the AMOC-slowdown scenario during the period when the global ocean is in an initial period of energy disequilibrium. We have, however, not addressed the possibility of fast oceanic teleconnections in our analysis. Given that in our study we use a simple thermodynamic ocean, it was not possible to assess how the oceanic response could have modified the picture. Coupled modeling studies, however, find similar tropical temperature and precipitation responses to high latitude cooling, and several studies have argued for the dominance of the atmospheric teleconnection in determining the global climate response (Chiang et al. 2008; Timmermann et al. 2005). In particular, study by Timmermann et al. (2007) shows that the AMOC weakening affects the northeastern tropical Pacific through changes in evaporation, mixing and Ekman forcing and further leads to ENSO intensification. Wu et al. (2008) attempt to distinguish between ocean and atmosphere driven changes, showing that the development of tropical Atlantic dipole is a consequence of the ocean dynamics while the tropical Pacific response comes from the atmosphere-surface ocean interactions.

We did not consider the role of extratropical eddy momentum fluxes that may be an important factor in determining how extratropical cooling alters the tropical Hadley circulation. Extratropical eddies have been shown to determine a sizable fraction of the Hadley cell mass flux over the areas where the Rossby number is small (Kim and Lee 2001; Walker and Schneider 2006). Observational data also suggest that the interannual variability in Hadley cell strength is strongly correlated to the eddy stress fluctuations with the stationary eddies playing the major role in the stronger winter Hadley cell variability and transients in the weaker summer cell (Caballero 2007; Caballero and Anderson 2009). The role that extratropical eddy momentum fluxes play in our problem has not yet been addressed, to our knowledge, and it remains a future task. Our brief analysis of the momentum budget during the transient adjustment to imposed high latitude cooling suggests that conditions before the tropical ITCZ shift are favorable for the more significant role of the extratropical eddy momentum fluxes. Further work is required to understand the dynamics of the adjustment period and the role of eddy momentum and heat fluxes in high to low latitude communication.


The authors acknowledge the support of Danish National Research Foundation, the Office of Science (BER), US Department of Energy (Award No. DE-FG02-08ER64588), and the National Science Foundation (Award No. ATM-0628628 and OCE-0902774). We thank Karen Shell and Alexandra Jonko for providing their radiative kernels and their assistance in its use, to Inez Fung on her invaluable comments and collaboration support and to Eigil Kaas and Peter L. Langen on their thoughtful suggestions. We are grateful to Andrew Friedman, Shih-Yu Lee, Yuwei Liu and Ching-Yee Chang on stimulating discussions.

Copyright information

© Springer-Verlag 2012