Abstract

The connection between midlatitude Atlantic sea surface temperature (SST) anomalies and tropical SST variations during boreal summer and fall are investigated using a coupled general circulation model (GCM). This research follows on an observational study that finds that, using linear inverse modeling (LIM), predictions of boreal summer tropical Atlantic Meridional Mode (AMM) variations can be made with skill exceeding persistence with lead times of about one year. The LIM framework identified extratropical Atlantic SST anomalies as important precursors to the AMM variations.

The authors have corroborated this finding using a general circulation model coupled to a slab ocean, which represents a completely different physical basis from the LIM. Initializing the GCM with the LIM-derived “optimal” SST anomaly in November results in a steady equatorward propagation of SST anomalies into the subtropics during the following boreal spring. Thereafter, the GCM suggests that two possible feedbacks propagate the SST anomalies farther equatorward and westward with minimal loss of amplitude: the dominant wind–evaporation–SST (WES) thermodynamic feedback and a secondary low-cloud–SST radiative feedback. This study shows that this result has strong seasonal dependence and consists of nonlinear interactions when considering warm and cold “optimal” conditions separately. One main finding is that oceanic dynamics are not essential to understanding extratropical–tropical interaction in the Atlantic basin. The authors also discuss the results of the study in context with previous studies investigating the extratropical forcing of tropical air–sea variability.

1. Introduction

There exist two dominant forms of coupled ocean–atmosphere variability in the tropical Atlantic basin [see Xie and Carton (2004) for a review]. The most dominant form, the Atlantic Meridional Mode (AMM; Chiang and Vimont 2004, hereafter CV04) is shown in Fig. 1 using regressions of sea surface temperature (SST) and 10-m horizontal wind onto CV04’s AMM index for boreal spring and fall. The AMM is a thermodynamically driven phenomenon that dominates SST variance in all seasons except boreal summer (Hastenrath and Heller 1977; Moura and Shukla 1981; Nobre and Shukla 1996; Ruiz-Barradas et al. 2000). The secondary mode, the Atlantic Niño, dominates SST variance during boreal summer and is characterized by a dynamical ocean–atmosphere interaction very similar to, albeit much weaker than, El Niño–Southern Oscillation (ENSO; Zebiak 1993; Carton and Huang 1994). The AMM has been linked to rainfall variability in Brazil during boreal spring (Hastenrath 1978) and the African Sahel during boreal summer (Folland et al. 1986). More recently, Vimont and Kossin (2007, hereafter VK07) showed that the AMM is well correlated to Atlantic tropical cyclone activity during boreal fall, mainly through the AMM’s modulation of large-scale environmental conditions in the main development region (MDR; Gray 1983; Goldenberg et al. 2001; Kossin and Vimont 2007, hereafter KV07; Smirnov and Vimont 2011). Meanwhile, the Atlantic Niño has been associated with Sahel rainfall and variability of the African easterly jet (Hsieh and Cook 2005; Nicholson and Webster 2007).

Fig. 1.

Regressions of SST (shaded) and surface horizontal wind (vectors) onto the AMM index from CV04 for (a) March–May and (b) August–October. The black contour encloses areas significant at the 98% confidence level using a two-tailed Student’s t test. Wind vectors are only plotted if they exceed the 98% confidence level.

Fig. 1.

Regressions of SST (shaded) and surface horizontal wind (vectors) onto the AMM index from CV04 for (a) March–May and (b) August–October. The black contour encloses areas significant at the 98% confidence level using a two-tailed Student’s t test. Wind vectors are only plotted if they exceed the 98% confidence level.

Although it is thought that the AMM requires external forcing for its existence, many studies have shown that the mode is destabilized (or at least made less stable) by a feedback between surface wind, evaporation, and SST (the WES feedback) that can maintain elevated SSTs after external forcing has subsided (Liu and Xie 1994; Chang et al. 1997; Xie 1999; Vimont 2010). The WES feedback seems to be confined to the tropics (Chang et al. 2000) and occurs when SST anomalies force an atmospheric response that reinforces the original SST anomalies through wind-induced latent heat flux (LHFLX) changes. Liu and Xie (1994) and Vimont (2010) show that the WES feedback induces an equatorward and westward propagation of coupled disturbances in the tropics. In theory, a thermodynamic mode like the AMM can exist in an atmosphere coupled to a static ocean. However, Servain et al. (1999) showed significant coherence between the time series of the meridional Atlantic ITCZ position (a proxy for the AMM) and the thermocline slope along the equator (a proxy for the Atlantic Niño), suggesting oceanic processes may be significant. Two other key issues that remain to be resolved are as follows: (i) the northern extent of the WES feedback phenomenon and (ii) the importance (relative to remote effects) of the WES feedback in sustaining and propagating SST anomalies into the deep tropics.

Many prior studies have investigated sources of tropical Atlantic variability and predictability and have shown the important roles of external forcing from El Niño–Southern Oscillation (ENSO; Curtis and Hastenrath 1995; Chang et al. 1997; Penland and Matrosova 1998; Chang et al. 1998; Ruiz-Barradas et al. 2000; Repelli and Nobre 2004) the North Atlantic Oscillation (NAO; Czaja et al. 2002; Kushnir et al. 2006), and the Atlantic multidecadal oscillation (AMO; Kushnir 1994; Mann and Park 1994; Kerr 2000; Enfield et al. 2001; Kerr 2005; VK07; KV07). While the NAO and ENSO connections operate on seasonal time scales, VK07 proposed that the AMM is also tied to the low-frequency variability of the AMO. VK07 show that the AMO (defined as the average SST anomaly over the North Atlantic basin north of 30°N) and AMM are well correlated (r = 0.60) but tend to exhibit higher correlations when the AMO leads the AMM by roughly one year. AMO variations likely exist because of a combination of processes including both tropical and extratropical air–sea heat exchange (Delworth and Greatbatch 2000), ocean circulation (Sutton and Hodson 2003; Knight et al. 2005), aerosol and dust forcing (Stendel et al. 2006) as well as potential anthropogenic contributions (Knight 2009); the relative contributions of each of these processes has not been determined.

The connection between the AMO and AMM that is hypothesized in VK07 and KV07 has been shown more explicitly in other studies. Zhang and Delworth (2005) used a fully coupled model to show that a large pulse of freshwater in the northern Atlantic Ocean (intended to induce a slowing of the thermohaline circulation) led to a southward shift in the intertropical convergence zone (ITCZ) within a decade. Chiang and Bitz (2005) used a slab ocean model to show that an increase in high latitude land and sea ice cover can also affect the ITCZ. Because a slab ocean cannot create anomalous oceanic dynamics, the authors concluded that a thermodynamic pathway (through the WES feedback) is responsible for the tropical circulation changes. By forcing an idealized atmospheric GCM coupled to a slab ocean with anomalous deep oceanic heat transport, Kang et al. (2008) and Kang et al. (2009) show that cloud-related radiative feedbacks can generate a tropical response through changes in atmospheric heat transport. This argument is based on a steady-state global energy budget and does not explicitly involve the WES feedback. Finally, Mann and Emanuel (2006) and Evan et al. (2009) argue that local, radiative (anthropogenic and natural) effects strongly influence tropical Atlantic variations and may explain the upward trend in tropical Atlantic SST during the last 30 years.

Research has shown that the tropical Atlantic SST response to ENSO and the NAO tends to exhibit maximum variance during boreal spring, as the oceanic mixed layer integrates ENSO- or NAO-related forcing through the boreal winter (Ruiz-Barradas et al. 2000; Czaja et al. 2002; Czaja 2004). Additionally, Mahajan (2008) explained that the WES feedback is more efficient in causing asymmetric SST anomalies when the ITCZ is close to the equator, as it is during boreal spring. Meanwhile, KV07 and Vimont (2012) suggest that during boreal summer and fall, AMM variations respond more to extratropical Atlantic forcing, rather than ENSO. Although variance of the AMM tends to peak during the boreal spring (CV04), the AMM does appear to vary during other seasons as well, as shown in Fig. 1. Note that the regression for boreal spring (Fig. 1a) resembles the pan-Atlantic “SST tripole” anomaly pattern (Tanimoto and Xie 1999; Czaja and Frankignoul 2002; Kushnir et al. 2006). Meanwhile, the pattern in late summer and fall (Fig. 1b) is markedly weaker in amplitude and resembles the North Atlantic “horseshoe” (Czaja and Frankignoul 2002). Another distinction between spring and fall is the lack of a tropical SST anomaly dipole during the latter. Indeed, prior studies have found similar results and shown that the anticorrelation between northern and southern tropical Atlantic SST anomalies is rather weak through most of the year (Houghton and Tourre 1992; Enfield et al. 1999).

KV07 and Vimont (2012) found that the boreal fall AMM index could be predicted with skill exceeding persistence from 11 months in advance, based on a statistical approach called linear inverse modeling (LIM). An in-depth discussion of the LIM technique can be found in Penland and Sardeshmukh (1995, hereafter PS95), where the authors apply the technique to ENSO prediction. Briefly, the LIM consists of maximizing the amount of explained lagged covariance of a given field, commonly SST anomalies, to produce “optimal” structures that maximally excite the dominant modes of variability. The LIM has previously been employed in the tropical Atlantic by Penland and Matrosova (1998). Vimont (2011) investigates the source of tropical Atlantic predictability in the LIM and finds that while tropical Pacific SST variations are essential for predicting boreal spring AMM variations (consistent with ENSO seasonality), mid- to high-latitude Atlantic initial conditions are particularly important for generating tropical Atlantic variations (using the LIM) in boreal fall. Although Vimont (2011) finds that the “optimal” structure includes a large SST anomaly signature in the extratropical Atlantic that transitions into an AMM-like state in the LIM framework, the study provides no physical explanation as to how the transition occurs. Other techniques are needed to ensure that the signal in the LIM is real, as opposed to a statistical artifact. In this study, we use a GCM to help decipher whether the LIM results can be corroborated using a different basis, and if so, the mechanism of how the SST signal is transferred from the midlatitudes in the tropics. Also, note that the “optimal” structure found by Vimont (2011) resembles the AMO pattern. Here we only address the tropical portion of the LIM result (pertaining to the AMM) and omit any discussion of conditions in the high-latitude (north of 50°N) Atlantic.

The outline for this paper is as follows. In section 2, we briefly introduce the LIM results of Vimont (2011) and explain the GCM setup that we used based on the LIM results. In section 3, we show that GCM simulations largely reproduce the LIM and develop an AMM-like signal during the boreal fall. In section 4, we provide a physical interpretation of how the tropical Atlantic may be excited by extratropical forcing and discuss the roles of seasonality and nonlinearity. Finally, conclusions and suggestions for future research are provided in section 5, along with a discussion of the limitations of our study.

2. Data and methods

a. Initial conditions from the LIM

Initial SST conditions for the model experiments are obtained from LIM analysis of SST anomalies over the Atlantic and Pacific basins (following Vimont 2011). LIM is described in more detail in PS95, but an overview is briefly presented here. Linear inverse modeling assumes that the tendency of the state of a system can be described by a first-order multivariate Markov model:

 
formula

where x is the “state” of the system (in this case SST anomalies over the region 30°S – 75°N, 120°–15°E), is the propagator matrix that quantifies the interaction between dominant modes of x, and ξ represents stochastic forcing. The solution to the homogeneous part of (1) can be written as

 
formula

In (2), is the “Green’s function”, which can be decomposed using singular value decomposition to yield matrices that contain initial conditions (columns in ; these are referred to herein as the “optimal” structures) that develop into final structures (columns in ) with amplitude growth corresponding to the associated singular values in Σ. The LIM analysis in this study follows that of Vimont (2011), and uses the National Centers for Environmental Prediction (NCEP) reanalysis (Kalnay et al. 1996) and Hadley (Rayner et al. 2006) SST datasets to ensure consistency. Here, we use monthly SST anomalies from 1950 to 2005 over the region 30°S–75°N, 120°–15°E. SST anomalies were found by removing the annual cycle, smoothing temporally with a 3-month running mean, and spatially with a 1–4-6–4-1 smoother. Before conducting the LIM, empirical orthogonal function (EOF)–principal component (PC) analysis is applied to the data as a prefilter, and enough PCs are retained to explain about 80% of the SST variance (about 18 EOFs). The is found using τ = 11 months, though the choice of τ, at least within a few months of 11, is not critical to the results. The amplitude of the initial condition is obtained by projecting the optimal structure onto the EOF-filtered SST anomalies to obtain a time series. The optimal structure is multiplied by two standard deviations of the time series, and as such the initial condition represents a two standard deviation anomaly of the associated optimal structure.

The initial conditions are taken from the second optimal structure (the first optimal is the precursor to ENSO; see PS95) as derived from the NCEP and Hadley based LIM analysis, as shown in Fig. 2. Note that for the imposed initial conditions, all anomalies outside of the Atlantic basin are “zeroed.” Both initial conditions in Fig. 2 show large loadings in the midlatitude Atlantic but strongly differ in the tropics. We will investigate the different responses to the midlatitude and tropical portions of the SST optimal, as described below. Figure 2 also shows the final SST structures into which the initial conditions develop (based on the LIM) with a lag of 11 months. Note that the final structures bear a resemblance to the AMM spatial structure for boreal summer (Smirnov and Vimont 2011; Vimont 2011).

Fig. 2.

LIM-derived SST (a),(c) initial conditions and (b),(d) final conditions derived as the second “optimal” structure (and final condition) from an 11-month Green’s function. The amplitude is representative of a two standard deviation anomaly for optimal initial condition during the August–October season. For comparison, results are shown from the (a),(c) NCEP reanalysis and from the (c),(d) HadISST data. The dashed line in (a),(c) denotes the 23°N cut-off between the “MID” and “LOW” initial conditions (see Table 1). For model simulations, SST anomalies are zeroed outside of the Atlantic and south of 30°S.

Fig. 2.

LIM-derived SST (a),(c) initial conditions and (b),(d) final conditions derived as the second “optimal” structure (and final condition) from an 11-month Green’s function. The amplitude is representative of a two standard deviation anomaly for optimal initial condition during the August–October season. For comparison, results are shown from the (a),(c) NCEP reanalysis and from the (c),(d) HadISST data. The dashed line in (a),(c) denotes the 23°N cut-off between the “MID” and “LOW” initial conditions (see Table 1). For model simulations, SST anomalies are zeroed outside of the Atlantic and south of 30°S.

b. Model description and setup

We use the National Center for Atmospheric Research (NCAR) Community Atmospheric Model version 3.1 [(CAM3.1); CAM is the atmospheric component of the Community Climate System Model version 3.1 (CCSM3.1); see Collins et al. (2006)]. The model is run using the default Eulerian dynamical core with T42 resolution (approximately 2.8° resolution in physical space), 26 vertical levels on a hybrid “sigma-pressure” vertical coordinate and a 20-min time step. The surface sensible and latent heat fluxes are calculated using a standard bulk formula that includes a stability-dependent transfer coefficient. The CAM model is coupled to a land and sea ice model and either a noninteractive data–ocean model (DOM, hereafter CAM+DOM) or an interactive slab-ocean model (SOM, hereafter CAM+SOM). The main difference between the SOM and DOM is that the SOM allows for thermodynamic ocean–atmosphere coupling while the DOM does not. Nonetheless, the DOM is a useful tool for investigating causal relationships.

Model simulations that are run for this analysis are described in Table 1. First, we ran two 40-yr-long control simulations using CAM+DOM (DOM.CLIMO) or CAM+SOM (SOM.CLIMO). The DOM.CLIMO simulation was forced with climatological SSTs, while the SOM.CLIMO simulation used a seasonally varying climatological ocean heat flux correction term [QFLUX; see  appendix A from Vimont et al. (2009) for a description of the CAM+SOM]. Note that for all runs involving CAM+SOM our mixed layer depth was temporally and spatially fixed at 50 m. We discuss the limitations of this assumption in section 5.

Table 1.

Simulations conducted using the CAM3.1 atmospheric GCM with either a prescribed SST field (DOM) or an interactive, slab ocean (SOM). See text for full description.

Simulations conducted using the CAM3.1 atmospheric GCM with either a prescribed SST field (DOM) or an interactive, slab ocean (SOM). See text for full description.
Simulations conducted using the CAM3.1 atmospheric GCM with either a prescribed SST field (DOM) or an interactive, slab ocean (SOM). See text for full description.

To investigate the evolution of the SST optimal structure shown in Fig. 2, we run ensemble model experiments using the CAM+SOM and 30 ensemble members (unless otherwise noted: see Table 1). Each ensemble member is initialized using the atmosphere, land, and oceanic (mixed layer heat content) initial conditions obtained from output of the SOM.CLIMO experiment. The oceanic initial conditions are altered by adding (subtracting) the LIM-derived anomaly shown in Fig. 2, hereafter referred to as the warm (cold) ensemble. Each ensemble member is initialized on 1 November and allowed to run for 13 months to cover the entirety of the following year’s fall season, as suggested by the LIM forecast skill. Three sets of warm and cold ensemble experiments are run initialized with (i) the full (SOM.TOTAL) Atlantic initial condition seen in Fig. 2, (ii) only the tropical portion (SOM.LOW) of the initial condition, and (iii) only the midlatitude (SOM.MID) portion of the initial condition. These three experiments were conducted with (i) the NCEP-derived initial condition (Fig. 2a) and (ii) the Hadley-derived initial condition (Fig. 2c). For the SOM.MID (SOM.LOW) simulations, anomalies south (north) of 23° were zeroed out, followed by two successive applications of a two-dimensional 1–2–1 spatial smoother to avoid sharp gradients.

Another set of ensemble experiments (DOM.MID) was motivated by a plethora of evidence (e.g., Table 1 in Kushnir et al. 2002) showing that the midlatitude atmospheric response to SST anomalies is very sensitive to seasonality (Peng et al. 1995). The DOM.MID experiment was similar to SOM.MID except SST anomalies were prescribed for the 10-yr duration of the simulation, to investigate how the atmospheric response to SST anomalies changes through the course of the annual cycle. Because of the fast decorrelation time (1–2 months) of midlatitude atmospheric fields, it was assumed that each year of the DOM-MID simulation represented an independent degree of freedom for statistical testing.

The results from the ensemble experiments are shown as the mean difference between the warm and cold ensemble members averaged over the three-month period November–January (NDJ), February–April (FMA), May–July (MJJ), and August–October (ASO). This method assumes that the warm and cold ensembles respond linearly and with opposite sign to the imposed forcing. As mentioned in Kushnir et al. (2002), this is a difficult assumption to make because of the strong sensitivity of the response on model depiction of the midlatitude atmosphere. With the availability of a control SOM simulation (SOM.CLIMO), we investigate the nonlinearity in the GCM simulations by separately comparing the warm and cold simulations to the control. Statistical significance is assessed using a two-tailed t test on the difference in ensemble means, with a sample size equal to the number of ensemble members. Grid points exceeding the 95% confidence level are assumed to be significant, unless otherwise noted. When necessary, we interpolated raw, monthly-averaged data from hybrid sigma-pressure coordinates to pure pressure coordinates using a logarithmic-linear interpolation.

3. Results

The response of the coupled CAM+SOM model to the optimal SST initial conditions is shown in Fig. 3 for NCEP-based LIM initial conditions and Fig. 4 for Hadley-based LIM initial conditions. The response to the full (tropical and midlatitude SSTs; SOM.TOTAL) Atlantic initial condition is shown in the left columns of Figs. 3 and 4. For the NCEP-based SOM.TOTAL (Fig. 3), initial SST anomalies in the midlatitude Atlantic (around 35°N) decay from late boreal fall (NDJ) through the following summer, while significant SST anomalies develop in the eastern tropical Atlantic (around 15°N, 30°–60°W) between late boreal winter (FMA) and the following summer (MJJ). These tropical anomalies are weak (between 0.2° and 0.4°C) but statistically significant. The Hadley-based initial conditions (Fig. 4) show a similar evolution, except that the midlatitude SST anomalies evolve very differently through the boreal winter (FMA), and the developing tropical SST anomalies during boreal summer (MJJ) are more widespread. The lack of ocean dynamics or external radiative forcing implies that these anomalies develop purely due to thermodynamically coupled processes. The development of these tropical anomalies is consistent with the evolution depicted in the LIM (Figs. 2b,d).

Fig. 3.

Evolution of SST anomalies for (left) SOM.TOTAL, (middle) SOM.MID, and (right) SOM.LOW forcing simulations using the NCEP-based initial condition. Each plot shows the mean difference between the warm and cold ensembles averaged over the 3-month periods (top to bottom) NDJ, FMA, MJJ, and ASO. The contour encloses the area that is significant at the 95%. See Table 1 for descriptions of each simulation.

Fig. 3.

Evolution of SST anomalies for (left) SOM.TOTAL, (middle) SOM.MID, and (right) SOM.LOW forcing simulations using the NCEP-based initial condition. Each plot shows the mean difference between the warm and cold ensembles averaged over the 3-month periods (top to bottom) NDJ, FMA, MJJ, and ASO. The contour encloses the area that is significant at the 95%. See Table 1 for descriptions of each simulation.

Fig. 4.

As in Fig. 3, but using the Hadley-based initial condition. The box in the top-right plot was used as the averaging region for Fig. 9.

Fig. 4.

As in Fig. 3, but using the Hadley-based initial condition. The box in the top-right plot was used as the averaging region for Fig. 9.

The LIM-based optimal initial conditions in Fig. 2 have substantial amplitude in the extratropical Atlantic but reduced amplitude in the tropical Atlantic. However, it is widely accepted that the tropical atmosphere is much more sensitive to small temperature perturbations (or gradients) than the midlatitude atmosphere (e.g., Hoskins and Karoly 1981). Thus, it is worth exploring whether the development of tropical SST anomalies in the SOM.TOTAL simulations can be traced to initial SST anomalies in the midlatitudes. or in the tropics, separately. This is shown in the middle and right columns in Figs. 3 and 4 for the midlatitude only (SOM.MID) and tropical only forcing (SOM.LOW), respectively.

The response in both the NCEP- and Hadley-based SOM.MID simulations (middle columns in Figs. 3 and 4) is consistent, with both simulations showing a gradual evolution of SST anomalies from the midlatitudes (in NDJ) into the subtropics and tropics as quickly as months 3–6 into the simulation (FMA). Of particular interest is the robust character of the tropical response, especially by late summer of year 1, where there is a maintenance of a large statistically significant area of positive SST anomalies of covering most of the tropical North Atlantic (along 15°N) in simulations with either set of initial conditions (Figs. 3 and 4, bottom rows). Note that this response is very similar to the LIM-derived final conditions in Figs. 2b,d, as well as the AMM SST spatial structures described in Nobre and Shukla (1996), Ruiz-Barradas et al. (2000), and CV04, all of which did not stratify data into seasons. The boreal summer and fall tropical response to the NCEP-based initial conditions (Fig. 3) is more robust than the response to the Hadley-based initial conditions (Fig. 4), though it is possible that differences may be due to limited sample size (30 ensembles) rather than fundamental differences in initial conditions. Note that throughout the simulations, high-latitude SST anomalies are maintained south of Greenland, making the northern Atlantic SST anomaly pattern similar to, though weaker, than the AMO (Knight et al. 2005). The difference in the mid- and high-latitude portions of the Atlantic between the GCM experiments and the LIM may be due to a lack of ice dynamics and associated oceanic convection in the CAM+SOM framework.

In addition to examining the mean difference in SST, Fig. 5 shows the difference in MDR (defined as 10°–20°N, 20°–60°W) SST (black) and SLP (gray) during ASO for each ensemble member of the NCEP-based SOM.MID simulation. The mean difference averaged across the ensembles is 0.33°C (−0.38 mb) and 24 (22) out of the 30 members have a higher (lower) MDR SST (SLP) in the warm simulation, compared to the cold. Note that the correlation between the SST and SLP differences is −0.76, which is strongly significant at the 99% confidence level and suggests a coupled process (Saravanan and Chang 1999). We have repeated Fig. 5 using the Hadley-based SOM.MID simulation and have found similar, though slightly less robust, results.

Fig. 5.

The difference in MDR SST (black) and SLP (gray) between each warm and cold ensemble during ASO from the NCEP-based SOM.MID simulation. The solid and dashed lines represent the average SST and SLP difference, respectively, across all ensemble members.

Fig. 5.

The difference in MDR SST (black) and SLP (gray) between each warm and cold ensemble during ASO from the NCEP-based SOM.MID simulation. The solid and dashed lines represent the average SST and SLP difference, respectively, across all ensemble members.

The tropical AMM-like response is not as clear in the SOM.LOW simulations (right columns of Figs. 3 and 4). Indeed, the SOM.LOW response to the NCEP-based initial conditions (Fig. 3, right column) actually has the opposite polarity as the NCEP-based SOM.TOTAL and SOM.MID simulations. Thus, if we consider the SOM.TOTAL response as the sum of the SOM.MID and SOM.LOW responses, we can see why the Hadley-based SOM.TOTAL experiment shows a stronger tropical signature during ASO compared to the NCEP-based SOM.TOTAL; this is due to the destructive interference from the initial tropical SST anomaly in NCEP-based LIM. The NCEP-based initial conditions exhibit SST anomalies on the order of 0.1°–0.3°C across the equatorial Atlantic, in a pattern that resembles the Atlantic Niño mode (Xie and Carton 2004). The subsequent development of a negative AMM-like signal in the following spring is consistent with findings of Okumura and Xie (2006). Further analysis is needed to address the potential relationship between the Atlantic Niño and the AMM. In contrast to the NCEP-based initial conditions, Hadley-based initial conditions (Fig. 2c) exhibit negligible anomalies in the tropics. The lack of a response in the Hadley-based SOM.LOW experiment (Fig. 4, right column) reflects this lack of sizeable tropical initial conditions. The lack of consistency between the NCEP- and Hadley-based initial conditions in the tropics suggests that tropical contribution to the optimal SST anomaly is less robust than the midlatitude contribution.

Figure 6 shows the evolution of the atmospheric response to the SST initial conditions. We have chosen to focus on the NCEP-based SOM.MID simulation (Fig. 3, middle column), as it seems to highlight the tropical response during the following year most vividly and is fairly consistent with both SOM.TOTAL experiments and the Hadley-based SOM.MID experiment. Figure 6 shows the same seasonal evolution as Fig. 3, except for SLP and lowest-model-level (∼992 hPa) horizontal wind. Comparing Fig. 6 with the middle column of Fig. 3 reveals that the midlatitude SST anomaly causes an overhead and slightly downstream lowering of surface pressure during NDJ, though it is interesting that only a small southern portion of this field is statistically significant at the 90% confidence level. In general, the finding of a downstream anomalous surface low-pressure is fairly consistent with both past theoretical (e.g., Hoskins and Karoly 1981; Webster 1981) and GCM studies (Palmer and Sun 1985; Kushnir and Held 1996). As noted by Kushnir et al. (2002), there is no consensus of how a midlatitude atmosphere should respond to an SST anomaly, as the answer depends on seasonality, spatial extent and, interestingly, on the GCM.

Fig. 6.

Evolution of SLP anomalies for NCEP-based SOM.MID averaged over the 3-month periods (a) NDJ, (b) FMA, (c) MJJ, and (d) ASO. Each plot shows the mean difference between the warm and cold ensembles. The contour encloses the area that is significant at the 90%.

Fig. 6.

Evolution of SLP anomalies for NCEP-based SOM.MID averaged over the 3-month periods (a) NDJ, (b) FMA, (c) MJJ, and (d) ASO. Each plot shows the mean difference between the warm and cold ensembles. The contour encloses the area that is significant at the 90%.

To investigate the vertical structure of the atmospheric response, Fig. 7 shows the mean difference of the temperature and geopotential height fields during NDJ for the NCEP-based SOM.MID simulation as a zonal cross section meridionally averaged from 32° to 40°N (roughly the center of the initial, anomalous SST). Note the SLP anomaly from Fig. 6 shows up as a low-level negative height anomaly between about 60° and 30°W and its vertical extent is limited to 600 hPa, while only the lowest 100 hPa is statistically significant at the 90% level. To the east, between about 30°W and 30°E, is a complex combination of a baroclinic and equivalent barotropic response that is statistically significant and compares well with linear theory (see Fig. 4a from Hoskins and Karoly 1981). Palmer and Sun (1985) and Ferranti et al. (1994) both obtained equivalent barotropic downstream high-pressure responses when forcing different GCMs with positive SST anomalies over the north Atlantic. Their results found the dependency of the downstream 500-hPa height on SST to be about 20 m K−1, while Fig. 7 suggests a smaller value of about 10 m K−1. Furthermore, note that the downstream barotropic high is about 30° east of the SST anomalies, compared to the 15° found by Palmer and Sun (1985). The GCM’s depiction of the quasi-stationary waves (e.g., the NAO) is one of the critical factors that determines the atmospheric response seen in Fig. 7 (Kushnir et al. 2002).

Fig. 7.

The response of the temperature (shaded) and geopotential height fields (contoured, interval 2 m) from the NCEP-based SOM.MID simulation averaged over NDJ. Values represent the mean difference between the warm and cold ensembles. The hatching indicates regions where the geopotential height field is significantly different from zero at the 90% confidence level using a two-tailed Student’s t test.

Fig. 7.

The response of the temperature (shaded) and geopotential height fields (contoured, interval 2 m) from the NCEP-based SOM.MID simulation averaged over NDJ. Values represent the mean difference between the warm and cold ensembles. The hatching indicates regions where the geopotential height field is significantly different from zero at the 90% confidence level using a two-tailed Student’s t test.

Figure 6 shows that the atmospheric response to the SST anomaly changes drastically from NDJ to FMA, when the accompanying negative SLP anomaly shifts west of the evolving positive tropical SST anomalies. This kind of response is akin to the stationary response in Gill (1980) and Hoskins and Karoly (1981), based on a linearization of the momentum and vorticity equations, who showed that the tropical atmosphere responds to diabatic heating with a surface low to the north and west of the heat source. However, the Gill model employed a beta-plane approximation with a Rossby radius of deformation of ∼10° latitude, while the anomalies seen during FMA of Fig. 6 are centered about 25° from the equator. Another uncertainty is that the Gill model, along with several other tropical circulation models, typically assumes that deep convection is the main atmospheric response to an SST anomaly. To follow up on this, Fig. 8 shows the mean difference in convective precipitation (PRECC) between the warm and cold ensembles of the NCEP-based SOM.MID simulation. The lack of convective precipitation anomalies during FMA suggests that the atmosphere may be forced through an alternate mechanism.

Fig. 8.

The response in the convective precipitation (PRECC, mm day−1) from the NCEP-based SOM.MID simulation averaged over (a) NDJ, (b) FMA, (c) MJJ, and (d) ASO. Values represent the mean difference between the warm and cold ensembles. The contour indicates regions that are significantly different from zero at the 95% confidence level using a two-tailed Student t test.

Fig. 8.

The response in the convective precipitation (PRECC, mm day−1) from the NCEP-based SOM.MID simulation averaged over (a) NDJ, (b) FMA, (c) MJJ, and (d) ASO. Values represent the mean difference between the warm and cold ensembles. The contour indicates regions that are significantly different from zero at the 95% confidence level using a two-tailed Student t test.

Transitioning into MJJ of Fig. 6, there is a general westward and equatorward expansion of statistically significant negative SLP anomalies, along with accompanying cyclonic circulation. There is also a dipole of precipitation anomalies in the tropics, indicative of a northward shift in the ITCZ (Fig. 8c). In addition, the surface wind anomalies superimposed on the climatological easterly winds result in a slackening of the wind speed, suggestive of the WES feedback. This anomalous pattern strengthens in amplitude during ASO (Figs. 6d and 8d) and strongly resembles a positive AMM in the SST, SLP, wind, and precipitation fields (Smirnov and Vimont 2011). The vertical pattern of the atmospheric response during the late boreal summer and fall of year 1 (not shown) has a baroclinic structure that is consistent with observational findings and model simulations of the AMM (Smirnov and Vimont 2011). The baroclinic circulation is associated with a reduction in wind shear that significantly overlaps with the MDR and has strong implications for an increase in tropical cyclone frequency and strength. In the next section, we discuss the physical mechanism of how the SST anomalies propagate from the midlatitudes into the tropics.

4. Physical mechanism

In section 3 it was shown that midlatitude Atlantic SST anomalies lead to a coupled response that can generate SST anomalies in the tropics. In this section we investigate the cause of the tropical SST anomalies (section 4a), and diagnose the midlatitude response to the initial midlatitude SST anomalies (section 4b).

a. Development of tropical SST anomalies

In the CAM+SOM model formulation, the development of SST anomalies can only occur via changes in the net surface heat flux. The evolution of SST anomalies in the NCEP-based and Hadley-based SOM.MID simulations is shown in the left and right columns of Fig. 9, respectively. The top panels of Fig. 9 show the time–latitude evolution of SST (shaded), net surface energy flux (for the CAM+SOM, the sum of shortwave, longwave, latent heat and sensible heat fluxes; contoured), and lowest-model-level winds as a zonally-averaged (40°–20°W; see top-right panel of Fig. 4 for a box) mean difference between the warm and cold ensembles in the SOM.MID experiments. All fluxes are defined to be positive upward (i.e., when heat is transferred from the ocean to the atmosphere). The NCEP-based plot (Fig. 9a) shows the gradual migration of positive SST anomalies from the region of large midlatitude initial conditions (30°–45°N, November) into the tropics and subtropics by the following boreal winter (February onward). A similar progression is seen in the Hadley-based experiment, though tropical SST anomalies develop about a month later, and do not achieve nearly the same intensity or southward extent as the NCEP-based experiment.

Fig. 9.

Evolution of SST (shaded), net surface energy flux (contoured, interval 2 W m−2) and surface wind (vectors) shown as a time–latitude section averaged from 40°–20°W (see box in top-right plot of Fig. 4) for the (a) NCEP-based and (b) Hadley-based SOM.MID simulation. Wind vectors are plotted if their magnitude exceeds the 90% significance level based on a two-tailed Student’s t test. Evolution of LHFLX (shaded) and net surface energy flux (contoured, interval 2 W m−2) for the (c) NCEP-based and (d) Hadley-based SOM.MID simulation.

Fig. 9.

Evolution of SST (shaded), net surface energy flux (contoured, interval 2 W m−2) and surface wind (vectors) shown as a time–latitude section averaged from 40°–20°W (see box in top-right plot of Fig. 4) for the (a) NCEP-based and (b) Hadley-based SOM.MID simulation. Wind vectors are plotted if their magnitude exceeds the 90% significance level based on a two-tailed Student’s t test. Evolution of LHFLX (shaded) and net surface energy flux (contoured, interval 2 W m−2) for the (c) NCEP-based and (d) Hadley-based SOM.MID simulation.

The net surface heat flux (by design) is consistent with the evolution of the SST field, with the atmosphere leading the ocean by a month or two given a 50-m mixed layer depth. Figures 9a,b show that the initial SST anomaly causes a rapid and intense negative atmospheric feedback, which eliminates all SST anomalies north of ∼40°N by January of year 1. Both simulations also show a relatively quick response in the tropics beginning in January (February) for the NCEP-based (Hadley-based) SOM.MID simulation. The earlier onset of tropical surface heat flux anomalies (by only one month) seems vital to maintaining SST anomalies in the tropics and reiterates the impact of seasonality because the NCEP-based SOM.MID simulation is able to propagate large SST anomalies farther south and more coherently than the Hadley-based SOM.MID simulation. In the next section, we show further evidence for the importance of seasonality by analyzing the atmospheric response when our “optimal” anomaly is prescribed throughout the course of an annual cycle (DOM.MID experiment). Inspection of the subtropical region (15°–25°N) from March to July reveals a positive air-sea feedback because the net energy flux is negative (into the ocean) over already elevated SSTs; this is much more pronounced in the NCEP-based SOM.MID simulation.

It has been shown that LHFLX variations tend to be the dominant term in the anomalous surface energy budget over windy areas of the ocean (Cayan 1992). To investigate the role of LHFLX variations, the bottom panels of Fig. 9 compare the response of the net surface energy flux (contoured; interval 2 W m−2) and the LHFLX (shaded) in the NCEP-based (Fig. 9c) and Hadley-based (Fig. 9d) SOM.MID simulation. In Figs. 9c,d, the anomalous LHFLX dominates the net surface energy flux in most areas throughout the course of the simulation. One exception is in the subtropical region (15°–23°N) during MJJ of Figs. 9c,d when there is a discrepancy between the LHFLX and net surface energy flux, which implies an important role for the radiative fluxes; this is explored later.

Now that we have established that anomalous LHFLX is the cause of initial tropical SST anomalies during FMA, Fig. 10 shows the decomposition of the LHFLX term into contributions from the zonal wind (LHFLXU; Fig. 10b) and vertical moisture gradient (LHFLXQ; Fig. 10c) component for the NCEP-based SOM.MID simulation (see  appendix for details). Figure 10 shows that LHFLXU dominates the total anomalous LHFLX over the southern portion of the SST anomalies during FMA. In accord with Fig. 10a, Fig. 10b shows that the initial subtropical SST anomalies are caused and maintained by LHFLXU, at least through April before temporarily abating, followed by a secondary, weaker resurgence during August and September (not shown). The signal from the moisture component is rather small over the areas where the wind component is large, which is inconsistent with Mahajan (2008), who found that in a climatological control run using CAM+SOM, the wind portion of the LHFLX was generally anticorrelated with the moisture gradient portion. The effectiveness of the WES-feedback in generating SST anomalies in the subtropics appears to be strongly dependent on the southward extent of the initial anomaly (cf. Figs. 2a,c in the region 20°–30°N, 20°–30°W). To confirm our hypothesis that the WES feedback is critical to equatorward SST anomaly propagation, we reran the NCEP-based SOM.MID experiment with fixed, climatological, seasonally varying surface wind speed over the tropical Atlantic Ocean (SOM.WES-OFF simulation; see Mahajan et al. 2009, for a description of the model setup). Note that the surface wind speed was only prescribed in the calculation of the latent and sensible heat flux. Figure 11 shows the evolution of the difference in SST anomalies between the warm and cold simulations of SOM.WES-OFF. Although there are significant regions in the tropics by ASO, the values are only ∼¼ of those in the SOM.MID simulation. Similarly, the convective precipitation response during ASO is roughly ∼⅓ of SOM.MID (not shown). Thus, the WES feedback is a dominant process of equatorward heat propagation, but the existence of significant regions of tropical SST anomalies during SOM.WES-OFF suggests a nondormant role for the moisture component of the LHFLX (radiative fluxes were negligible, not shown). These results compare well with the recent model simulations of Mahajan et al. (2011), but further research is needed to assess if this can be found in observations.

Fig. 10.

The decomposition of the (a) total LHFLX into a (b) zonal wind and (c) specific humidity gradient component (see  appendix). Values represent the mean difference between the warm and cold ensembles of the NCEP-based SOM.MID simulation during FMA. The contour represents areas that are significant at the 95% confidence level based on a two-sided Student’s t test.

Fig. 10.

The decomposition of the (a) total LHFLX into a (b) zonal wind and (c) specific humidity gradient component (see  appendix). Values represent the mean difference between the warm and cold ensembles of the NCEP-based SOM.MID simulation during FMA. The contour represents areas that are significant at the 95% confidence level based on a two-sided Student’s t test.

Fig. 11.

Evolution of SST anomalies for SOM.WES-OFF forcing simulations using the NCEP-based initial condition. Each plot shows the mean difference between the warm and cold ensembles averaged over the 3-month periods (a) NDJ, (b) FMA, (c) MJJ, and (d) ASO. The contour encloses the area that is significant at the 95%.

Fig. 11.

Evolution of SST anomalies for SOM.WES-OFF forcing simulations using the NCEP-based initial condition. Each plot shows the mean difference between the warm and cold ensembles averaged over the 3-month periods (a) NDJ, (b) FMA, (c) MJJ, and (d) ASO. The contour encloses the area that is significant at the 95%.

b. The role of seasonality and nonlinearity in midlatitude air–sea coupling

As mentioned in the previous section and in Kushnir et al. (2002, see their Table 1), seasonality is thought to play a critical role in the atmospheric response to midlatitude SST anomalies. To probe at this question further, we ran the uncoupled DOM.MID experiment that prescribed the same SST anomaly as in Fig. 2 (NCEP initial conditions) for the 10-yr duration of the simulation. Analysis of the mean difference in the SLP and LHFLX fields for NDJ and FMA is shown in Fig. 12. Note that there are significant differences in the atmospheric response between the two seasons. During NDJ (Fig. 12a), the atmosphere responds with a surface low pressure directly over the SST anomaly, similar to the NDJ response from the coupled simulation (Fig. 6a). Furthermore, the LHFLX is generally negative in the subtropics and tropics, which is, again, qualitatively consistent with the coupled experiment results in Fig. 9. On the contrary, in FMA (Fig. 12b), there are no negative LHFLX anomalies in the subtropics and the pattern resembles a general negative atmospheric feedback that quickly damps the SST anomaly. The reason for such a stark difference in responses is still unclear but may involve an evolving projection onto the NAO; this issue will not be addressed within this manuscript.

Fig. 12.

The mean difference between the warm and cold polarity runs of the DOM.MID simulation is shown for SLP (contoured, interval 0.5 mb) and latent heat flux (positive upward; shaded) averaged over (a) NDJ and (b) FMA. Also shown are surface horizontal wind vectors only they are significant at the 90% confidence level.

Fig. 12.

The mean difference between the warm and cold polarity runs of the DOM.MID simulation is shown for SLP (contoured, interval 0.5 mb) and latent heat flux (positive upward; shaded) averaged over (a) NDJ and (b) FMA. Also shown are surface horizontal wind vectors only they are significant at the 90% confidence level.

It has been shown that the midlatitude atmosphere does not respond linearly to changes in surface boundary conditions (Frankignoul 1985; Kushnir et al. 2002; Deser et al. 2004, hereafter D04). To check for linearity in the midlatitude response, we separately compared the mean of warm and cold ensembles of the NCEP-based SOM.MID simulations with the control (SOM.CLIMO) during NDJ and ASO, shown in Fig. 13 for SLP and SST. Note that the results in Fig. 13 were calculated by subtracting the control run from the warm ensembles (WARM-CLIM; Figs. 13a,c) and subtracting the cold ensembles from the control run (CLIM-COLD; Figs. 13b,d). Thus, in theory, a linear response would generate the same amplitude and polarity in the structures shown in Fig. 13. During NDJ (Figs. 13a,b), there is a significant nonlinear component of the response because the WARM-CLIM component develops an intense negative NAO-like structure in the SLP field and dampens a portion of the original SST anomaly through mainly the latent and sensible heat fluxes (not shown). In contrast, the CLIM-COLD has a lack of significant SLP anomalies directly over the SST anomaly and generates a downstream high pressure with a weak projection onto the positive NAO. However, it appears that seasonality once again may be important because during FMA the WARM-CLIM SLP response shows no projection onto the NAO, while the CLIM-COLD does show a modest, though insignificant, projection onto the negative NAO (not shown). Further research is needed to address if this seasonal dependence is robust.

Fig. 13.

The mean difference between the NCEP-based (a),(c) warm ensembles and SOM.CLIMO and (b),(d) SOM.CLIMO and cold ensembles for SST (shaded) and SLP (contoured) for (a),(b) NDJ (SLP contour interval 0.25-hPa) and (c),(d) ASO (SLP contour interval 0.1). The hatching indicates areas of SST that are significant at the 95% level. For SLP, positive (negative) contours are solid (dashed) and the zero contour is thickened. Note that the SLP north of 40°N is omitted for ASO due to noise.

Fig. 13.

The mean difference between the NCEP-based (a),(c) warm ensembles and SOM.CLIMO and (b),(d) SOM.CLIMO and cold ensembles for SST (shaded) and SLP (contoured) for (a),(b) NDJ (SLP contour interval 0.25-hPa) and (c),(d) ASO (SLP contour interval 0.1). The hatching indicates areas of SST that are significant at the 95% level. For SLP, positive (negative) contours are solid (dashed) and the zero contour is thickened. Note that the SLP north of 40°N is omitted for ASO due to noise.

The nonlinearity of the response of CAM3.1 to extratropical SST anomalies was thoroughly investigated by D04, who decomposed the response into a direct and indirect component. The indirect component is the result of the SST anomaly inducing an atmospheric response that projects onto the intrinsic modes of variability, whereas the direct component is the localized atmospheric response, and mathematically, the residual between the full and indirect responses. Even though our initial anomalies were 10°–15° farther south than D04, Fig. 13 suggests our results are in agreement with D04 in showing that the WARM-CLIM projects much more strongly onto NAO than CLIM-COLD. To investigate the linear contribution of the thermodynamic response to the SST anomaly, Fig. 14 shows the total diabatic heating rate QDIAB as a longitude–pressure cross section meridionally averaged from 32° to 40°N (the highest-amplitude regions from the initial conditions) for the WARM-CLIM and CLIM-COLD components during NDJ. We see a vertically coherent structure of QDIAB centered roughly over the SST anomaly in both the WARM-CLIM and CLIM-COLD components. The QDIAB response in WARM-CLIM is significant at the 95% mainly between 400 and 600 hPa. Meanwhile, the CLIM-COLD component has two significant areas, one associated with boundary layer processes and one with an apparent reduction in deep convection. Above 900 hPa, the dominant term contributing to QDIAB was from condensation processes (not shown), which can be interpreted as anomalous deep convection and cloud processes (Nigam et al. 2000). We note that D04 had a similar finding where the WARM-CLIM component extends higher in the troposphere, but they also found a full column response, while our finding on shows significant anomalies in WARM-CLIM mainly above 700 hPa.

Fig. 14.

A longitude–pressure cross section of the mean difference in the total diabatic heating rate QDIAB between the (a) warm ensembles and SOM.CONTROL and (b) SOM.CONTROL and the cold ensembles from the NCEP-based SOM.MID simulation. Values were temporally averaged over the NDJ months and meridionally averaged from 32° and 40°N (to coincide with the highest amplitudes in Fig. 2). The black contour shows areas that are significant at the 95% confidence level.

Fig. 14.

A longitude–pressure cross section of the mean difference in the total diabatic heating rate QDIAB between the (a) warm ensembles and SOM.CONTROL and (b) SOM.CONTROL and the cold ensembles from the NCEP-based SOM.MID simulation. Values were temporally averaged over the NDJ months and meridionally averaged from 32° and 40°N (to coincide with the highest amplitudes in Fig. 2). The black contour shows areas that are significant at the 95% confidence level.

Focusing once again on Fig. 13, we see that by ASO (Figs. 13c,d), the SST response between the WARM-CLIM and CLIM-COLD components is surprisingly similar in depicting tropical SST anomalies, and akin to the structure of the AMM (see Fig. 1). Inspection of the monthly resolved evolution of SST anomalies (not shown) suggests that persistence into May–June is critical for equatorward propagation during ASO. Further evidence of this can be seen by re-examining the response of convective precipitation (PRECC), previously shown in Fig. 8. During FMA (Fig. 8b), there are large SST anomalies in the eastern subtropical Atlantic (Fig. 3, middle column), but there is a lack of any response in PRECC. However, by MJJ and especially ASO, the PRECC response amplifies and suggests a northward displaced ITCZ position, yet the magnitude of the SST anomalies off the northwest coast of Africa remains virtually unchanged from FMA to MJJ (cf. middle column of Fig. 3 to Figs. 8c,d). The difference in the PRECC response from FMA to MJJ given a similar SST anomaly pattern strongly suggests that the mean background state is vital to sustaining SST anomalies. The anomalously displaced ITCZ is in line with previous studies of tropical excitation through extratropical SST or ice forcing (Chiang and Bitz 2005; Zhang and Delworth 2005; Chiang et al. 2008). However, one distinction between our work and previous work on extratropical–tropical connection is that the significant SLP and SST anomalies that we observe are essentially limited to the Atlantic basin (not shown), whereas the aforementioned studies show a hemisphere-wide response. Also, we note that in our case it takes about 6 months before the ITCZ begins to shift, which is faster than the referenced studies, probably because our initial forcing is closer to the equator.

Next, we investigate the aforementioned discrepancy between the LHFLX and net surface heat flux during MJJ. As seen in Fig. 9, the anomalous LHFLX is initially responsible for driving SST anomalies into the tropical region. However, during MJJ of year 1, both the NCEP-based and Hadley-based SOM.MID simulations suggest the LHFLX anomalies are weak in the tropics (see Fig. 9c,d). An inspection of the remaining terms in the net surface energy budget suggests that radiative flux anomalies briefly assume dominant roles in maintaining SST anomalies in the subtropical and tropical Atlantic during MJJ. Figure 15 compares the mean difference of the net surface heat flux to the shortwave radiative flux and the LHFLX in the NCEP-based SOM.MID simulation during MJJ and reveals that there are no significant LHFLX anomalies in the tropics. However, there is a strong anomalous shortwave radiative flux response off the northwest coast of Africa collocated with the SST anomalies at that time (see Fig. 3, middle column, for MJJ period), suggesting the possibility of a low-cloud–SST feedback (Klein et al. 1995). Figure 16 shows the average low-cloud fraction (contours) during MJJ over the Atlantic sector from International Satellite Cloud Climatology Project (ISCCP) climatology over 1981–2007 (Rossow and Schiffer 1999); note the tongue of higher values in the eastern portion of the North Atlantic Ocean. One complicating finding is that CAM3.1 has a very large bias in low cloud during MJJ, shown in Fig. 16 as the difference between ISCCP and SOM.CLIMO climatology (shading). The area of the largest bias in low-cloud fraction shown in Fig. 16 strongly overlaps with the area of strong downward net surface heat flux in Fig. 15. Further research is currently being done to investigate how this bias may contribute to a positive low-cloud–SST feedback, and its subsequent influence on the equatorward propagation of SST anomalies during the critical early summer months.

Fig. 15.

The mean difference in the (a) total surface energy flux, (b) the shortwave radiative flux, and (c) the LHFLX averaged for the MJJ months. Values show the mean difference between warm and cold ensembles from the NCEP-based SOM.MID simulation. Contoured areas are significant at the 90% confidence level.

Fig. 15.

The mean difference in the (a) total surface energy flux, (b) the shortwave radiative flux, and (c) the LHFLX averaged for the MJJ months. Values show the mean difference between warm and cold ensembles from the NCEP-based SOM.MID simulation. Contoured areas are significant at the 90% confidence level.

Fig. 16.

The bias in low-cloud cover percentage (shaded, %) of the SOM.CLIMO simulation, compared to ISCCP climatology from 1981–2007 for MJJ (contoured; interval 10%).

Fig. 16.

The bias in low-cloud cover percentage (shaded, %) of the SOM.CLIMO simulation, compared to ISCCP climatology from 1981–2007 for MJJ (contoured; interval 10%).

5. Conclusions and discussion

a. Conclusions

We have investigated how midlatitude SST anomalies can force tropical Atlantic air–sea variability on interseasonal time scales. The motivation for this work came from a finding by KV07 and Vimont (2011) that used a LIM to show that the AMM may be predictable up to a year in advance but only during boreal summer and fall. To corroborate this finding using an alternate physical basis, we designed a GCM experiment that was based on the “optimal” pattern found by the LIM (Fig. 2). The GCM, which was coupled to a slab ocean, reproduced the LIM finding that initial midlatitude Atlantic SST anomalies in November generate an AMM-like response in the tropical Atlantic by the following boreal summer and fall. First, the midlatitude atmosphere responds to the positive SST anomalies by developing a downstream surface low pressure, which in the warm ensembles is consistent with past theoretical arguments using linear theory (Hoskins and Karoly 1981). This anomalous cyclonic circulation quickly erodes the northern edge of the initial SST anomalies through an enhanced LHFLX. However, the same circulation also develops SST anomalies in the subtropical north Atlantic due to a relaxation of the mean easterly winds. Farther southward and equatorward propagation occurs through the WES feedback, which appears to operate up to about 25°N. During late boreal spring and early summer a positive feedback between low clouds and SST maintains the tropical SST anomalies. There is some uncertainty with this feedback given the bias shown by the GCM in low cloud cover during boreal summer (Fig. 16). By late boreal summer, the WES feedback re-emerges as the dominant signal carrier given an increasingly favorable background state as the ITCZ marches northward during the annual cycle. The WES feedback slowly shuts off around October and an enhanced LHFLX finally attenuates the remaining signal.

In addition to our main finding that the LIM result can be replicated by a GCM coupled to a slab ocean, we have done several sensitivity tests to shed light on various aspects of the extratropical–tropical interaction over the Atlantic. By repeating our experiment with prescribed (uncoupled) SSTs, we showed that seasonality plays a vital role in our main finding. When forcing with the same “optimal” SST anomaly during February, no equatorward propagation is seen. We attribute the strong impact of seasonality to the evolving background state in the north Atlantic, which features a southward shift of the jet stream from November to February. Many studies have investigated the atmospheric response to SST anomalies in the vicinity of a strong jet [see Kushnir et al. (2002) for a summary], but the results have been inconsistent and even conflicting, providing an area for further research.

We have investigated the linearity of the response by running both warm and cold simulations in addition to a control CAM+SOM simulation. Our results compare well with the result of D04, who showed that the warm ensembles project strongly onto the NAO while the cold ensembles do not. In accord with D04, we find that diabatic heating in the warm ensembles extends to a higher level in the troposphere compared with the cold simulations. This suggests the importance of deep convection in the atmospheric response; however, modeling deep convection with a GCM is prone to many simplifications and assumptions. Even though our SST anomaly is in the vicinity of the warm Gulf Stream extension region of the Atlantic Ocean, which is a climatologically favored area for deep convection, it may be presumptuous to make conclusions regarding the nonlinear response of deep convection to the SST anomaly.

b. Discussion

Although we have shown that a midlatitude SST anomaly in the Atlantic can propagate equatorward on interseasonal time scales through thermodynamic processes, we need to address several limitations of our result. First, we use a constant (50 m) mixed layer depth, whereas it is well known that the midlatitude ocean mixed layer depth experiences a strong seasonal cycle with amplitudes exceeding 150 m (Alexander and Deser 1995). Using a 50-m seasonally and spatially invariant mixed layer may be somewhat unrealistic, yet it is also a safe strategy because the climatological mixed layer depth during November in the mid-Atlantic basin is between 75 and 100 m (see Fig. 3 in Alexander and Deser 1995). Thus, the heat content anomaly that we impose is fairly conservative and certainly within the realm of natural variability. Second, note from Fig. 5, that five of the six cases where the cold ensemble MDR SST was higher than the warm ensemble occur in the middle 10 ensembles, suggesting a role for the low-frequency variability in the SOM.CLIMO simulation. Indeed, Fig. 17 shows the time series of MDR-averaged SST anomalies (bars: raw, line: 12-month running mean) during ASO from SOM.CLIMO simulation. Note that the ensemble numbers (1–30) correspond to initial conditions taken from November of years 10–39 of the SOM.CLIMO simulation, thus the ASO period during model years 11–40 corresponds to ASO of ensembles 1–30. The propensity of the low-frequency variability in the SOM.CLIMO simulation to exert an influence on the SOM.MID simulation suggests that each ensemble member is not entirely independent from its neighbor, as we have earlier assumed. Furthermore, the correlation between the unfiltered SOM.CLIMO ASO SST in the MDR (bars in Fig. 17) and the SST difference from Fig. 5 is −0.37 (significant at the 95% level), indicating nonlinearity in the sense that our initial condition had a smaller impact in separating a given warm and cold ensemble if the tropical Atlantic was already evolving toward a warm state. Third, we have not addressed the source of the initial SST anomaly in the midlatitudes. The long time scale of extratropical SST anomalies (Vimont 2011) suggests that oceanic processes such as variations in the Atlantic meridional overturning circulation or Gulf Stream may play an important role (Sutton and Allen 1997; Minobe et al. 2008). However, regardless of the source or time scale of the initial SST anomalies, our results suggest that the response can be carried to the tropics solely via thermodynamically coupled feedbacks. Fourth, we have not explained the high-latitude differences between the GCM simulations and the LIM “final” structure. Although both the NCEP- and Hadley-based SOM.MID simulations generate SST anomalies north of 50°N, these are very weak and misplaced compared to the LIM, preventing us from making any conclusions in this area, though providing a direction for future research.

Fig. 17.

The time series of ASO SST anomalies averaged over the MDR from the last 30 years of the SOM.CLIMO simulation. The bars indicate unfiltered, seasonal averages and the black line denotes seasonal averaging after smoothing with a 12-month running mean.

Fig. 17.

The time series of ASO SST anomalies averaged over the MDR from the last 30 years of the SOM.CLIMO simulation. The bars indicate unfiltered, seasonal averages and the black line denotes seasonal averaging after smoothing with a 12-month running mean.

Finally, it is also interesting to compare our results with past studies of midlatitude forcing of tropical variability. Our results are consistent the conclusions of Chiang and Bitz (2005) in showing that extratropical thermal forcing can affect the ITCZ position through thermodynamic processes alone, even though our anomaly was ∼15° farther south. Our results support those of Mahajan (2008) because we have demonstrated that the WES feedback is very important to the propagation of SST anomalies into the subtropics and tropics, south of 25°N. However, our results from the SOM.WES-OFF experiment are also similar to the recent work of Mahajan et al. (2011), which shows that extratropical excitation of tropical SST anomalies can occur in the absence of the WES feedback, albeit with greatly reduced amplitude. Additionally, we have showed that seasonality has a large impact on whether or not tropical excitation occurs, which has been omitted in past studies that focused on longer time scales. Although we focus on interseasonal time scales and our model does not have ocean dynamics and thus cannot be directly compared to the study by Zhang and Delworth (2005), we believe it is highly unlikely that a change in the strength of the AMOC can propagate SST anomalies into the tropics via oceanic processes in the short time scales that we have considered. Thus, we can conclude that oceanic dynamics are not essential in extratropical-to-tropical climate interactions in the North Atlantic.

Acknowledgments

We thank Salil Mahajan for discussions regarding the WES-OFF experiment. We would also like to thank three anonymous reviewers for very constructive suggestions regarding this manuscript. This work was funded by the National Science Foundation Grant ATM-0849689.

APPENDIX

Decomposition of Latent Heat Flux into a Wind and Specific Humidity Component

The bulk formula for calculating the LHFLX is

 
formula

where ρA is the air density (1.2 kg m−3), CE is a stability-dependent unitless surface drag coefficient (assumed to be constant at 1.5 × 10−3 for simplicity), L is the latent heat of vaporization (2.50e6 J kg−1), qsat is the saturated specific humidity at the sea surface (SST) or at the 2 m air temperature TA, and RH is the specific humidity at 2 m. The U is the lowest-model level wind speed, estimated as

 
formula

where u and υ are the zonal and meridional wind speeds that were taken directly from model output, respectively, and w* represents a gustiness parameter (see Czaja et al. 2002) that was found empirically as the difference between the 1000 hPa total wind speed and horizontal wind speed from NCEP reanalysis using monthly data from 1950 to 2008 only over the northern Atlantic. The gustiness parameter w* is approximated by the function

 
formula

where lat is the latitude in degrees. To estimate the zonal wind (specific humidity gradient) contribution to the total LHFLX, the specific humidity gradient (low-level wind speed U) was set to the climatological value from the SOM.CONTROL simulation. For example, the zonal wind speed contribution to the LHFLX is

 
formula

where UW (and UC) are found by setting the meridional wind speed υ to its climatological value from the SOM.CONTROL simulation:

 
formula
 
formula

REFERENCES

REFERENCES
Alexander
,
M. A.
, and
C.
Deser
,
1995
:
A mechanism for the recurrence of wintertime midlatitude SST anomalies
.
J. Phys. Oceanogr.
,
25
,
122
137
.
Carton
,
J.
, and
B.
Huang
,
1994
:
Warm events in the tropical Atlantic
.
J. Phys. Oceanogr.
,
24
,
888
903
.
Cayan
,
D. R.
,
1992
:
Latent and sensible heat flux anomalies over the northern oceans: The connection to monthly atmospheric circulation
.
J. Climate
,
5
,
354
369
.
Chang
,
P.
,
J.
Link
, and
H.
Li
,
1997
:
A decadal climate variation in the tropical Atlantic Ocean from thermodynamic air–sea interactions
.
Nature
,
385
,
516
518
.
Chang
,
P.
,
L.
Ji
,
H.
Li
,
C.
Penland
, and
L.
Matrosova
,
1998
:
Prediction of tropical Atlantic sea surface temperature
.
Geophys. Res. Lett.
,
25
,
1193
1196
.
Chang
,
P.
,
R.
Saravanan
,
L.
Ji
, and
G. C.
Hegerl
,
2000
:
The effect of local sea surface temperatures on the atmospheric circulations over the tropical Atlantic sector
.
J. Climate
,
13
,
2195
2216
.
Chiang
,
J. C. H.
, and
D. J.
Vimont
,
2004
:
Analagous Pacific and Atlantic meridional modes of tropical atmosphere–ocean variability
.
J. Climate
,
17
,
4143
4158
.
Chiang
,
J. C. H.
, and
C. M.
Bitz
,
2005
:
Influence of high latitude ice cover on the marine intertropical convergence zone
.
Climate Dyn.
,
25
,
477
496
,
doi:10.1007/s00382-005-0040-5
.
Chiang
,
J. C. H.
,
W.
Cheng
, and
C. M.
Bitz
,
2008
:
Fast teleconnections to the tropical Atlantic sector from Atlantic thermohaline adjustment
.
Geophys. Res. Lett.
,
35
,
L07704
,
doi:10.1029/2008GL033292
.
Collins
,
W. D.
, and
Coauthors
,
2006
:
The Community Climate System Model version 3 (CCSM3)
.
J. Climate
,
19
,
2122
2143
.
Curtis
,
S.
, and
S.
Hastenrath
,
1995
:
Forcing of anomalous sea surface temperature evolution in the tropical Atlantic during Pacific warm events
.
J. Geophys. Res.
,
100
,
15 835
15 847
.
Czaja
,
A.
,
2004
:
Why is the north tropical Atlantic SST variability stronger in boreal spring?
J. Climate
,
17
,
3017
3025
.
Czaja
,
A.
, and
C.
Frankignoul
,
2002
:
Observed impact of Atlantic SST anomalies on the North Atlantic Oscillation
.
J. Climate
,
15
,
606
623
.
Czaja
,
A.
,
P.
van der Vaart
, and
J.
Marshall
,
2002
:
A diagnostic study of the role of remote forcing in tropical Atlantic variability
.
J. Climate
,
15
,
3280
3290
.
Delworth
,
T. L.
, and
R. J.
Greatbatch
,
2000
:
Multidecadal thermohaline circulation variability driven by atmospheric surface flux forcing
.
J. Climate
,
13
,
1481
1495
.
Deser
,
C.
,
G.
Magnusdottir
,
R.
Saravanan
, and
A.
Phillips
,
2004
:
The effects of North Atlantic SST and sea ice anomalies on the winter circulation in CCM3. Part II: Direct and indirect components of the response
.
J. Climate
,
17
,
877
889
.
Enfield
,
D. B.
,
A. M.
Mestas-Nunez
,
D. A.
Mayer
, and
L.
Cid-Serrano
,
1999
:
How ubiquitous is the dipole relationship in the Tropical Atlantic sea surface temperatures?
J. Geophys. Res.
,
104
,
7841
7848
.
Enfield
,
D. B.
,
A. M.
Mestas-Nunez
, and
P. J.
Trimble
,
2001
:
The Atlantic multidecadal oscillation and its relation to rainfall and river flows in the continental U.S
.
Geophys. Res. Lett.
,
28
,
2077
2080
.
Evan
,
A. T.
,
D. J.
Vimont
,
R.
Bennartz
,
J. P.
Kossin
, and
A. K.
Heidinger
,
2009
:
The role of aerosols in the evolution of tropical North Atlantic Ocean temperature
.
Science
,
324
,
778
781
.
Ferranti
,
L.
,
F.
Molteni
, and
T. N.
Palmer
,
1994
:
Impact of localized tropical and extratropical SST anomalies in ensembles of seasonal GCM integrations
.
Quart. J. Roy. Meteor. Soc.
,
120
,
1613
1645
.
Folland
,
C. K.
,
T. N.
Palmer
, and
D. E.
Parker
,
1986
:
Sahel rainfall and worldwide sea temperatures, 1901–1985
.
Nature
,
320
,
602
607
.
Frankignoul
,
C.
,
1985
:
Sea surface temperature anomalies, planetary waves, and air–sea feedback in the middle latitudes
.
Rev. Geophys.
,
23
,
357
390
.
Gill
,
A. E.
,
1980
:
Some simple solutions for heat-induced tropical circulation
.
Quart. J. Roy. Meteor. Soc.
,
106
,
447
462
.
Goldenberg
,
S.
,
C. W.
Landsea
,
A.
Mestas-Nunez
, and
W. M.
Gray
,
2001
:
The recent increase in Atlantic hurricane activity: Causes and implications
.
Science
,
293
,
474
479
.
Gray
,
W. M.
,
1983
:
Atlantic seasonal hurricane frequency. Part I: El Niño and 30-mb quasi-biennial oscillation influences
.
Mon. Wea. Rev.
,
112
,
1649
1668
.
Hastenrath
,
S.
,
1978
:
On modes of tropical circulation and climate anomalies
.
J. Atmos. Sci.
,
35
,
2222
2231
.
Hastenrath
,
S.
, and
L.
Heller
,
1977
:
Dynamics of climate hazards in Northeast Brazil
.
Quart. J. Roy. Meteor. Soc.
,
103
,
77
92
.
Hoskins
,
B. J.
, and
D. J.
Karoly
,
1981
:
The steady linear response of a spherical atmosphere to thermal and orographic forcing
.
J. Atmos. Sci.
,
38
,
1179
1196
.
Houghton
,
R. W.
, and
Y. M.
Tourre
,
1992
:
Characteristics of low-frequency sea surface temperature fluctuations in the Tropical Atlantic
.
J. Climate
,
5
,
765
771
.
Hsieh
,
J.-S.
, and
K. H.
Cook
,
2005
:
Generation of African easterly wave disturbances: Relationship to the African easterly jet
.
Mon. Wea. Rev.
,
133
,
1311
1327
.
Kalnay
,
E.
, and
Coauthors
,
1996
:
The NCEP/NCAR 40-Year Reanalysis Project
.
Bull. Amer. Meteor. Soc.
,
77
,
437
471
.
Kang
,
S. M.
,
I. M.
Held
,
D. M. W.
Frierson
, and
M.
Zhao
,
2008
:
The response of the ITCZ to extratropical thermal forcing: Idealized slab-ocean experiments with a GCM
.
J. Climate
,
21
,
3521
3532
.
Kang
,
S. M.
,
D. M. W.
Frierson
, and
I. M.
Held
,
2009
:
The tropical response of extratropical thermal forcing in an idealized GCM: The importance of radiative feedbacks and convective paramaterization
.
J. Atmos. Sci.
,
66
,
2812
2827
.
Kerr
,
R. A.
,
2000
:
A North Atlantic climate pacemaker for the centuries
.
Science
,
288
,
1984
1986
.
Kerr
,
R. A.
,
2005
:
Atlantic climate pacemaker for millenia past, decades hence?
Science
,
309
,
41
42
.
Klein
,
S. A.
,
D. L.
Hartmann
, and
J. R.
Norris
,
1995
:
On the relationship among low-cloud stucture, sea surface temperature, and atmospheric circulation in the summertime northeast Pacific
.
J. Climate
,
8
,
1140
1155
.
Knight
,
J. R.
,
2009
:
The Atlantic multidecadal oscillation inferred from the forced climate response in coupled general circulation models
.
J. Climate
,
22
,
1610
1625
.
Knight
,
J. R.
,
R. J.
Allan
,
C. K.
Folland
,
N.
Vellinga
, and
M.
Mann
,
2005
:
A signature of persistent natural thermohaline circulation cycles in observed climate
.
Geophys. Res. Lett.
,
32
,
L20708
,
doi:10.1029/2005GL024233
.
Kossin
,
J. P.
, and
D. J.
Vimont
,
2007
:
A more general framework for understanding Atlantic hurricane variability and trends
.
Bull. Amer. Meteor. Soc.
,
88
,
1767
1781
.
Kushnir
,
Y.
,
1994
:
Interdecadal variations in North Atlantic sea surface temperature and associated atmospheric conditions
.
J. Climate
,
7
,
141
157
.
Kushnir
,
Y.
, and
I. M.
Held
,
1996
:
Equilibrium atmospheric response to North Atlantic SST anomalies
.
J. Climate
,
9
,
1208
1220
.
Kushnir
,
Y.
,
W. A.
Robinson
,
I.
Blade
,
N. M. J.
Hall
,
S.
Peng
, and
R.
Sutton
,
2002
:
Atmospheric GCM response to extratropical SST anomalies: Synthesis and evaluation
.
J. Climate
,
15
,
2233
2256
.
Kushnir
,
Y.
,
W. A.
Robinson
,
P.
Chang
, and
A. W.
Robertson
,
2006
:
The physical basis for predicting Atlantic sector seasonal-to-interannual climate variability
.
J. Climate
,
19
,
5949
5970
.
Liu
,
Z.
, and
S.
Xie
,
1994
:
Equatorward propagation of coupled air–sea disturbances with application to the annual cycle of the eastern tropical Pacific
.
J. Atmos. Sci.
,
51
,
3807
3822
.
Mahajan
,
S.
,
2008
:
Free and forced tropical variability: Role of the wind–evaporation—sea surface temperature (WES) feedback
.
Ph.D. thesis. Texas A&M University, 150 pp
.
Mahajan
,
S.
,
R.
Saravanan
, and
P.
Chang
,
2009
:
The role of the wind-evaporation-sea surface temperature (WES) feedback in air-sea coupled tropical variability
.
Atmos. Res.
,
94
,
19
36
.
Mahajan
,
S.
,
R.
Saravanan
, and
P.
Chang
,
2011
:
The role of the wind–evaporation–sea surface temperature (WES) feedback as a thermodynamic pathway for the equatorward propagation of high-latitude sea ice–induced cold anomalies
.
J. Climate
,
24
,
1350
1361
.
Mann
,
M. E.
, and
J.
Park
,
1994
:
Global-scale modes of surface temperature variability on interannual to century timescales
.
J. Geophys. Res.
,
99
,
25 819
25 833
.
Mann
,
M. E.
, and
K.
Emanuel
,
2006
:
Atlantic hurricane trends linked to climate change
.
Eos, Trans. Amer. Geophys. Union
,
24
,
233
244
.
Minobe
,
S.
,
A.
Kuwano-Yoshida
,
N.
Komori
,
S.-P.
Xie
, and
R. J.
Small
,
2008
:
Influence of the Gulf Stream on the troposphere
.
Nature
,
452
,
206
209
.
Moura
,
A. D.
, and
J.
Shukla
,
1981
:
On the dynamics of droughts in northeast Brazil: Observations, theory and numerical experiments with a general circulation model
.
J. Atmos. Sci.
,
38
,
2653
2675
.
Nicholson
,
S. E.
, and
P. J.
Webster
,
2007
:
A physical basis for the interannual variability of rainfall in the Sahel
.
Quart. J. Roy. Meteor. Soc.
,
133
,
2065
2084
.
Nigam
,
S.
,
C.
Chung
, and
E.
DeWeaver
,
2000
:
ENSO diabatic heating in ECMWF and NCEP–NCAR reanalyses, and NCAR CCM3 simulation
.
J. Climate
,
13
,
3152
3171
.
Nobre
,
P.
, and
J.
Shukla
,
1996
:
Variations of sea surface temperature, wind stress, and rainfall over the tropical Atlantic and South America
.
J. Climate
,
9
,
2464
2479
.
Okumura
,
Y.
, and
S.-P.
Xie
,
2006
:
Some overlooked features of tropical Atlantic climate leading to new Niño-like phenomenon
.
J. Climate
,
19
,
5859
5874
.
Palmer
,
T. N.
, and
Z.
Sun
,
1985
:
A modeling and observational study of the relationship between sea surface temperature in the northwest Atlantic and the atmospheric general circulation
.
Quart. J. Roy. Meteor. Soc.
,
111
,
947
975
.
Peng
,
S.
,
L. A.
Mysak
,
H.
Ritchie
,
J.
Derome
, and
B.
Dugas
,
1995
:
The differences between early and midwinter atmospheric responses to sea surface temperature anomalies in the northwest Atlantic
.
J. Climate
,
8
,
137
157
.
Penland
,
C.
, and
P. D.
Sardeshmukh
,
1995
:
Optimal growth of tropical sea surface temperature anomalies
.
J. Climate
,
8
,
1999
2024
.
Penland
,
C.
, and
L.
Matrosova
,
1998
:
Prediction of tropical Atlantic sea surface temperatures using linear inverse modeling
.
J. Climate
,
11
,
483
496
.
Rayner
,
N. A.
,
P.
Brohan
,
D. E.
Parker
,
C. K.
Folland
,
J. J.
Kennedy
,
M.
Vanicek
,
T.
Ansell
, and
S. F. B.
Tett
,
2006
:
Improved analyses of changes and uncertainties in sea surface temperature measured in situ since the mid-nineteenth century: The HadSST2 data set
.
J. Climate
,
19
,
446
469
.
Repelli
,
C. A.
, and
P.
Nobre
,
2004
:
Statistical prediction of sea surface temperature over the tropical Atlantic
.
Int. J. Climatol.
,
24
,
45
55
.
Rossow
,
W. B.
, and
R. A.
Schiffer
,
1999
:
Advances in understanding clouds from ISCCP
.
Bull. Amer. Meteor. Soc.
,
80
,
2261
2287
.
Ruiz-Barradas
,
A.
,
J. A.
Carton
, and
S.
Nigam
,
2000
:
Structure of interannual-to-decadal climate variability in the tropical Atlantic sector
.
J. Climate
,
13
,
3285
3297
.
Saravanan
,
R.
, and
P.
Chang
,
1999
:
Oceanic mixed layer feedback and tropical Atlantic variability
.
Geophys. Res. Lett.
,
26
,
3629
3632
.
Servain
,
J.
,
I.
Wainer
,
J. P.
McCreary
, and
A.
Dessier
,
1999
:
Relationship between the equatorial and meridional modes of climatic variability in the tropical Atlantic
.
Geophys. Res. Lett.
,
26
,
485
488
.
Smirnov
,
D.
, and
D. J.
Vimont
,
2011
:
Variability of the Atlantic meridional mode during the Atlantic hurricane season
.
J. Climate
,
24
,
1409
1424
.
Stendel
,
M.
,
I. A.
Mogensen
, and
J. H.
Christensen
,
2006
:
Influence of various forcings on global climate in historical times using a coupled atmosphere–ocean general circulation model
.
Climate Dyn.
,
26
,
1
15
.
Sutton
,
R. T.
, and
M. R.
Allen
,
1997
:
Decadal predictability of North Atlantic sea surface temperature and climate
.
Nature
,
388
,
563
567
.
Sutton
,
R. T.
, and
D. L. R.
Hodson
,
2003
:
Influence of the ocean on North Atlantic climate variability 1871–1999
.
J. Climate
,
16
,
3296
3313
.
Tanimoto
,
Y.
, and
S.-P.
Xie
,
1999
:
Ocean–atmosphere variability over the pan-Atlantic basin
.
J. Meteor. Soc. Japan
,
77
,
31
46
.
Vimont
,
D. J.
,
2010
:
Transient growth of thermodynamically coupled disturbances in the tropics under an equatorially symmetric mean state
.
J. Climate
,
23
,
5771
5789
.
Vimont
,
D. J.
,
2012
:
Analysis of the Atlantic meridional mode using linear inverse modeling: Seasonality and regional influences
.
J. Climate
,
25
,
1194
1212
.
Vimont
,
D. J.
, and
J. P.
Kossin
,
2007
:
The Atlantic meridional mode and hurricane activity
.
Geophys. Res. Lett.
,
34
,
L07709
,
doi:10.1029/2006GL029683
.
Vimont
,
D. J.
,
M.
Alexander
, and
A.
Fontaine
,
2009
:
Midlatitude excitation of tropical variability in the Pacific: The role of thermodynamic coupling and seasonality
.
J. Climate
,
22
,
518
534
.
Webster
,
P.
,
1981
:
Mechanisms determining the atmospheric response to SST anomalies
.
J. Atmos. Sci.
,
38
,
554
571
.
Xie
,
S.-P.
,
1999
:
A dynamic ocean–atmosphere model of the tropical Atlantic decadal variability
.
J. Climate
,
12
,
64
70
.
Xie
,
S.-P.
, and
J. A.
Carton
,
2004
:
Tropical Atlantic variability: Patterns, mechanisms, and impacts
.
Earth’s Climate: The Ocean–Atmosphere Interaction, Geophys. Monogr., Vol. 147, Amer. Geophys. Union, 121–142
.
Zebiak
,
S. E.
,
1993
:
Air–sea interaction in the equatorial Atlantic region
.
J. Climate
,
6
,
1567
1586
.
Zhang
,
R.
, and
T. L.
Delworth
,
2005
:
Simulated tropical response to a substantial weakening of the Atlantic thermohaline circulation
.
J. Climate
,
18
,
1853
1860
.