1 Introduction

The Atlantic meridional overturning circulation (AMOC)is a key element of the global ocean conveyor belt (Lozier,2010; Srokosz and Bryden, 2015). Changes in the AMOC can significantly influence the oceanic and atmospheric climate system (Wan et al., 2009; Mu et al., 2011), because the AMOC transports large amounts of heat and water through buoyancy-forced deeper-ocean thermohaline circulation and wind-driven upper-ocean circulation (Wen et al., 2011; Men et al., 2016). Previous climate change records indicate that abrupt climate transitions between glacial and interglacial conditions are associated with pronounced changes in the AMOC (McManus et al., 2004).Moreover, the AMOC has been suggested as the underlying ocean circulation responsible for the Atlantic multidecadal oscillation (AMO; Knight et al., 2005) that affects Atlantic hurricane activity (Landsea et al., 1999), Sahelian rainfall, and summer climate conditions over much of North America and Europe (Sutton and Hodson, 2005).

Even though the importance of the AMOC to Earth’s climate has been recognized for some time now, our understanding of the physical mechanisms that govern AMOC variability and its relationship to long-term climate variability remains incomplete partly because of the lack of long-term ocean observations. Direct continuous measurements of the AMOC have only become available since 2004 (Cunningham et al., 2007; Srokosz and Bryden, 2015). Although climate models simulate the broad structure of the AMOC, they disagree on the strength and temporal variability of the AMOC. Only a subset of models used in the Intergovernmental Panel on Climate Change(IPCC) fourth assessment simulates AMOC strength within the range of observationally based estimates during the late twentieth century. However, AMOC changes in response to global warming as predicted by these models vary considerably, ranging from nearly no change to an approximately 50% reduction in the strength of the overturning circulation (Schmittner et al., 2005). The disparity among the models reflects the complexity of AMOC dynamics and a lack of fundamental understanding of its maintenance and variability mechanisms.

Current understanding indicates that the AMOC is sustained primarily by the wind that provides an energy source for diapycnal mixing of heat in the interior ocean,driving the slow upwelling of the deep water into the thermocline (Wunsch, 2002). The source of the deep water is the subpolar North Atlantic, where oceanic deep convection controls the deep-water formation. Both diapycnal mixing and deep convection are highly complex processes that depend not only on the density structure and ambient flow condition of the ocean but also on atmospheric forcing conditions. For example, the deep convection in the Irminger Sea is closely related to the onset of the Greenland tip jet, a low-level atmospheric jet that is less than 2˚ wide and is strongly influenced by winter storm activities in the vicinity of Greenland (Pickart et al.,2003). The onset of deep convective events in the Labrador Sea is strongly affected by cold air outbreaks associated with synoptic weather systems during boreal winter(Våge et al., 2009). Strong wind events, such as those associated with tropical cyclones, have been suggested as an agent that promotes enhanced turbulent mixing in the oceans (Sriver and Huber, 2007).

制造能力的并集并不是简单的制造能力相加，而是涉及到制造活动在执行过程中的转移，在制造活动的转移过程中将影响制造服务的能力。多个制造活动集合的制造能力的逻辑关系如图4所示。

While recent studies have begun to unravel the complex interactions between large-scale ocean dynamics and small-scale oceanic eddies in determining the AMOC structure (Lozier, 2010), current ocean modeling studies estimate that including daily fluctuating fluxes can increase AMOC strength (Beena and von Storch, 2009) and that wind forcing, rather than buoyancy forcing, is the primary contributor to high-frequency fluctuations of the AMOC (Biastoch et al., 2008). However, Beena and von Storch (2009) used an empirical flux model, which is a linear stochastic model fitted to the daily output of a coupled general circulation model (CGCM) on a coarse grid resolution. As such, the empirical model may not resolve synoptic winter storm events. Therefore, although these ocean model studies shed new light on the role of atmospheric variability in the AMOC, the extent to which its impact on the climate system as a whole has not been addressed with a higher-spatial-resolution CGCM, because feedbacks between the ocean and atmosphere have not been fully taken into consideration.

We conduct a set of twin present-climate simulations by using the Community Climate System Model version 3 (CCSM3) developed at the National Center for Atmospheric Research. CCSM3 used in this study is referred to as T85x1 CCSM3, which consists of the Community Atmosphere Model and the Community Land Model using a spectral truncation of T85 (grid spacing of 1.4˚) coupled to the Parallel Ocean Program and the Community Sea Ice Model on a grid with approximately 1˚ horizontal resolution. More detailed information about CCSM3 can be found at the CCSM website at http://www.ccsm.ucar.edu/models. The performance of the CCSM in simulating present and past climate can be found in the special issue of the Journal of Climate (Vol. 19, No. 11).

We first present the structure and variability of the simulated AMOC. The AMOC strength is estimated by the maximum AMOC streamfunction within 20˚–50˚N and 0–2000 m. Fig.1 compares the time mean AMOC streamfunction and the time series of the streamfunction maximum between the CTRL and IE runs. Two notable findings emerge: first, the mean AMOC strength decreases from 22 Sv (1 Sv = 106 m3 s−1) in the CTRL to 17 Sv in the IE run (Figs.1a and 1c), accompanied by a mean reduction in the northward heat transport (NHT) within the Atlantic sector by approximately 0.2 PW (Fig.1b). Second, the CTRL simulation exhibits a strong and unambiguous decadal oscillation, as indicated by a statistically significant spectral peak near 16–20 years and an amplitude of approximately 4.5 Sv; this finding is consistent with a previous study (McManus et al., 2004). This decadal variability is significantly weakened in the IE run,and the corresponding spectral peak does not survive a significant test at 95% confidence (Fig.1d). The energetic AMOC variability produces a NHT variability of approximately 0.12 PW, which represents 10% of its mean maximum.

巧妇难为无米之炊，实施通识教育需要建设一批丰富的微课、慕课、精品资源共享课等网络资源。这些资源可以通过以下几种做法进行建设：一是学校可以通过选择并加入一个网络平台共享已经开发的课程资源，如超星泛雅平台，它是集慕课及精品课程建设、教学互动、资源管理、教学成果展示、教学管理评估于一体的新一代网络教学平台，专注于通识教育课程研发，平台上汇集了国内知名学者已经开发的涉及多个范畴的通识课程资源；二是选择或购买部分课程纳入学院的选修课程管理系统；三是任课教师自主开发建设部分基于慕课平台通识课程。另外，也要充分利用校外资源，高职院校之间有着鲜明的学科差异性和较强的互补性。

2 Model Descriptions and Experimental Design

More importantly, a study of whether and how the variability of the AMOC depends on the high-frequency component of the forcing is crucial, and the underlying processes of this forcing are not yet fully understood.Numerical experiments presented by Delworth and Greatbatch (2000) show that the multi-decadal AMOC variability is driven by the low-frequency surface heat flux forcing, and the high frequencies of the surface forcing contribute minimally to the AMOC variability. Presumably, the multi-decadal oscillation signal is contained in the low-frequency surface heat flux. However, the experiment does not explain the source of the multi-decadal oscillation signal in the surface heat flux. One possible scenario is that stochastic weather variability forces a redspectrum oceanic response, of which the multi-decadal oscillation is a part. The surface heat fluxes then record the multi-decadal oscillation in sea surface temperature(SST) through passive air-sea coupling. Therefore, when the ocean-only model is forced with the low-frequency surface heat fluxes, it gives back multi-decadal oscillation to the ocean. If this situation is true, then it means that the multi-decadal oscillation is ultimately sustained by synoptic weather variability. Unfortunately, the uncoupled standalone ocean general circulation model approach used by Delworth and Greatbatch (2000) is unsuitable for fully testing this hypothesis, thereby necessitating a new approach that can filter out synoptic weather variability within the coupled model framework.

The change in AMOC variability is also reflected in the simulated North Atlantic SST, as shown by a comparison of a modeled AMO index. Fig.1e shows that the AMO SST variance decreased significantly at time scales longer than 2 years. In particular, the statistically significant spectral peak near 20 years in the CTRL, which corresponds to the decadal variability of the AMOC, drops below the 95% confidence level in the IE. AMOC-associated SST changes are shown in Fig.2. Fig.2a shows an SST anomaly composite, which is derived by taking the difference between the average of SST anomalies in high and low AMOC states in the CTRL run. Maximum SST anomalies clearly form along the Gulf Stream extension region in the North Atlantic, with peak-to-peak values reaching as large as 6–7℃. Although the modeled SST anomaly does not exactly resemble the observed SST anomalies associated with the AMO, a warming (cooling)in the North Atlantic and a cooling (warming) in the South Atlantic exist, which correspond to a high (low)AMOC state. Fig.2b shows the mean SST difference between the IE and CTRL simulations. Consistent with the NHT reduction (Fig.1b), much of the North Atlantic cools with a maximum cooling of 3˚–4˚ along the Gulf Stream extension region, and the South Atlantic warms with a maximum warming of 1˚ off the coast of Africa. The dipole-like SST response in the Atlantic is consistent with the model SST response to a weakened AMOC (Fig.2). A direct comparison between the AMOC-induced SST anomaly (Fig.2a) in the control simulation and the mean SST change in the IE simulation (Fig.2b) reveals a similar dipole-like SST pattern in the Atlantic (expect for a different sign). This pattern suggests that the Atlantic SST change in the IE simulation is likely caused by the weakened AMOC due to the suppression of synoptic weather variability.

数据在采集和传输过程中会产生一些干扰，因此有必要对采集的数据进行预处理。如果采集值处于上下限之间，则作为正常数据处理，如果超出正常的分布范围，则表明信号受到干扰，此时应将其限定在合理的范围内。具体公式如下：

The novelty of the IE approach is its use of multiple realizations (six were used in this study) of the atmospheric component coupled to a single realization of other component models, including ocean, sea-ice, and land surface models. By contrast, the CTRL run uses only one realization of the atmospheric component coupled to a single realization of other component models. To elucidate the concept of IE, we express atmospheric surface flux (heat, momentum, and freshwater) anomalies as Q=QS+QI, where QS is the component driven by SST anomalies, and QI is the component induced by internal atmospheric dynamics. All the ensemble members are forced by the identical SST anomaly; thus, QS remains the same for all the ensemble members. However, QI varies from dif-ferent ensemble members, because each ensemble member starts from a slightly different atmospheric initial condition and the atmosphere is sensitive to its initial state.Therefore, ensemble averaging by IE at each coupled time step effectively weakens QI while keeping QS unaltered. As a result, flux anomalies passed to the ocean have a relatively larger contribution from QS in an IE run than in a standard CTRL run. Given the infinitely large ensemble number, the surface flux anomalies contain only the SST-induced component QS. In this experimental design, the ocean is forced by the portion of wind stress,surface heat, and freshwater flux anomalies, which are driven solely by SST anomalies. If the forced oceanic response is to sustain or even strengthen the existing SST anomalies, then a co-varying ocean-atmosphere pattern emerges in the IE run. By definition, this pattern of variability is considered an ocean-atmosphere coupled mode.Therefore, the IE approach is a powerful tool for extracting the ocean-atmosphere signal in a complex climate system model.

The analysis presented in this study is based on the monthly mean output of the numerical simulations. Each experiment of CCSM3 consists of 400-year integrations,which started with exactly the same oceanic initial condition. We analyze and compare the last 160 years of the 400-year simulations.

3 Results

1）使用酒精擦拭纸擦拭管道表面，移除任何残余的灰尘和其他污染物。不能使用丙酮、丁酮或其他不适于擦拭准备就绪的端头表面的溶液。

Fig.1 (a) 160-year time mean AMOC streamfunction from the CCSM3 CTRL (contour) and the difference of the time mean AMOC streamfunctions between IE and CTRL simulations. Contour interval is 2 Sv. (b) Zonally averaged Atlantic Ocean heat transport in CTRL (blue) and IE (red). (c) Time series of the maximum of AMOC streamfunctions from CTRL (blue) and IE (red) runs. (d) Spectra of the AMOC time series shown in c from CTRL (blue) and IE (red) run. (e)Spectra of AMO index defined by averaging SST over the North Atlantic from 0˚ to 75˚N and from 75˚W to 10˚W in CTRL (blue) and IE (red). In (d) and (e), the dashed curves denote significant test for spectral peaks at 95% level.

The simulations we performed can be categorized into the following runs: in one simulation, we use the standard coupling procedure, which is referred to hereafter as the CTRL run, to provide a baseline for other experiments; in the other simulation, we adopt a novel coupling procedure called interactive ensemble (IE; Kirtman et al., 2009),which is designed to significantly reduce the effect of weather variability generated by internal atmospheric dynamics on the coupling between the atmosphere and ocean. We refer to this run as the IE run in this paper.

To address this issue, we adopt a novel coupling technique and conduct a set of twin numerical experiments to assess the effect of weather variability generated by internal atmospheric dynamics on the coupling between the atmosphere and ocean. The remainder of this paper focuses primarily on the region of the Atlantic Ocean and is organized as follows: in Section 2, we outline our numerical model framework and introduce the novel coupling procedure and the two simulations to be compared;in Section 3, we analyze the model output data in detail and present the main findings; in Section 4, we summarize our findings and discuss their implications.

Fig.2 (a) SST anomaly associated with the AMOC oscillation in CCSM control simulation. The figure shows an SST composite derived by taking the difference between the average of SST anomalies in high and low AMOC states. Contour interval is 0.2℃. The star indicates the location where the buoyancy frequency time series is computed as shown in Fig.3. (b) 160-year time mean SST difference between the IE and control simulations. Contour interval is 0.5℃.

Fig.5 shows lag correlations between the first principal component time series of the March mean MLD and AMOC index in the CTRL and IE simulations. Statistically significant correlations (＞0.5) are observed in the CTRL run when MLD leads the AMOC by 3–5 years, but no statistically significant correlations are observed in the IE run at any lag. This finding implies that the deep convection shows a pronounced decadal variation that matches well with the decadal variability in the AMOC streamfunction index when the former leads the latter by approximately 3–5 years, thereby suggesting that in CCSM3, the deep convection is closely linked to AMOC variability, as shown in other studies (Danabasoglu, 2008;Kwon and Frankignoul, 2012). By contrast, the IE simulation displays no statistically significant decadal variation in either the deep convection or the AMOC time series, as well as no statistically significant lag correlations between the deep convection and AMOC variation (Fig.5).Unlike a previous approach that used a low-pass filtered surface fluxes from a coupled ocean-atmosphere model simulation to force an ocean model to diagnose multidecadal AMOC variability in the coupled model (Delworth and Greatbatch, 2000), the IE approach regards weather variability generated by internal atmospheric dynamics as a white-spectrum noise and filters the noise at all time scales within a coupled model framework. As such, the finding supports the notion that weather variability acts as an important stochastic forcing for the AMOC through its effect on deep convection in the subpolar North Atlantic (Kwon and Frankignoul, 2012).

One probable cause for the significantly weaker AMOC in the IE simulation is that the ensemble averaging employed by the IE procedure substantially reduces strong latent and sensible surface heat flux events accompanied by individual winter storms and cold air outbreaks, thereby weakening the deep convection activity and deep- water production that feeds to the AMOC. This effect is clearly revealed by the monthly mean time series of the buoyancy frequency and the turbulent heat fluxes calculated at a model deep convection site (46˚W, 54˚N) within the Labrador Sea (Fig.3). Fig.4 shows the winter mean mixed layer depth (MLD) in CTRL and the MLD difference between the IE and CTRL runs. The winter mean is computed by averaging each January-February-March(JFM) of the last 5 years of the 20-year segment from the CTRL and IE runs shown in Fig.3. In the CTRL run, the vertical mixing associated with winter deep convective events can penetrate to a depth of 2000 m. By contrast, in the IE run, the mixing depth is significantly reduced and limited to less than 1000 m (Fig.4). A major reduction of 31% (about 87 W m−2) in the boreal winter surface turbulent heat fluxes accompanies the reduced MLD in the IE run; this reduction is consistent with the reduced winter storm activity.

Fig.3 (a) Monthly mean time series of the net surface heat fluxes at at 46˚W, 54˚N in a 20-year segment of the CTRL (blue)and IE (red) simulation. The difference between the two heat fluxes primarily occurs in the winter season. The blue and red dashed lines in a indicate the winter mean (JFM) surface heat fluxes for CTRL and IE simulation, respectively. (b)Monthly mean time series of buoyancy frequency at the same location in the CTRL simulation. Near-zero value of buoyancy frequency marked by dark blue indicates strong vertical mixing. (c) The corresponding 20-year segment of the maximum AMOC streamfunction in the CTRL (blue) and IE (red) simulations. The dashed lines indicate the mean AMOC transport values. (d) Same as b except for the IE run.

Fig.4 Mean MLD in CCSM3 CTRL simulation (contour) and difference between MLD in the IE and CTRL runs (color).Contour interval is 200 m. The star indicates the location where the buoyancy frequency time series is computed as shown in Fig.3.

Fig.5 a) Lag correlations between the first principal component of March mean MLD and AMOC index in CTRL. b)Same as 1) except for IE run. The dashed lines indicate 95% significance level.

4 Summary and Discussion

In this study, we use coupled numerical simulations to illustrate the effect of synoptic weather variability on the strength and variability of the AMOC. Results suggest that the variability in the surface forcing due to weather significantly affects the strength and variability of the AMOC. With the use of a novel coupling technique, the strength and the variability of the AMOC are reduced when the weather variability is removed from the surface forcing. This finding implies that synoptic weather variability is important in driving the AMOC and its variability. If synoptic weather variability can considerably affect the AMOC as suggested by the IE run, then it may have contributed to AMOC changes during the second half of the twentieth century, as indicated by a poleward shift and increased storm track activities (Chang, 2007) and extreme storm events, particularly over the North Atlantic Ocean (Shaman et al., 2010). Our results imply that the intensified Atlantic weather variability during the twentieth century could have strongly influenced the AMOC’s long-term change by counteracting the effect as a result of increased subpolar ocean surface buoyancy in response to global warming. However, determining the effect of the changing synoptic weather variability on the AMOC by using direct observations currently presents a great challenge due to the lack of long-term direct measurements of the AMOC.

The strengthening of the AMOC seemingly contradicts the notion that global warming should cause the AMOC to decrease, because warming-induced changes in surface heat and freshwater flux in high latitudes can result in increased buoyancy at the ocean surface, thereby reducing deep convection in subpolar seas (Schmittner et al.,2005). Indeed, all IPCC models predict, albeit in different degrees, decreases in the AMOC strength under the SRES A1B scenario (Schmittner et al., 2005). The current study does not aim to debate whether the AMOC has strengthened or weakened over the twentieth century. Rather, it seeks to call attention to the active role of the atmospheric weather variability in the AMOC and related climate variability and change. While global warming has caused changes in subpolar ocean surface buoyancy, it has also altered synoptic-scale weather variability, both of which can affect the AMOC and associated climate variability and climate change. A recent study also suggests that AMOC changes may affect the Atlantic storm track (Woollings et al., 2012).

2.2 两组体质量净增值的比较 出生至42 d、42 d～2个月、2～3个月，体质量净增值干预组均大于对照组，差异有统计学意义(均P<0.05)，而3～6个月两组体质量净增值差异无统计学意义(P>0.05)。见表2。

Together, these recent findings indicate a potential feedback between the storm track and AMOC. Therefore,our discussion on long-term AMOC changes should consider not only changes in oceanic conditions but also changes in atmospheric weather variability. The strengthening in the storm track activity may have had such a dominant influence on the AMOC over the twentieth century that the weakening of the AMOC caused by the increase in the high-latitude ocean surface buoyancy may be outweighed by the strengthening of the AMOC driven by intensified storm track activity. If this idea is proven correct, then a continuous intensifying storm track activity,as projected by IPCC climate scenarios for the twenty-first century (Yin, 2005; Bengtsson et al., 2010), will likely impose an important constraint on future changes of the AMOC. As global warming continues, the mechanism and the degree to which the AMOC will change in the future will depend on the competition between the rates of the subpolar ocean surface buoyancy increase and storm track intensification. Future improvements of climate models need to focus not only on large- scale circulations but also on high-frequency, synoptic-scale weather features such as storm tracks because they are an integral part of the climate system.

Acknowledgements

This research is funded by the National Natural Science Foundation of China (Nos. 41276013, 41576004,41776009 and U1406401). The authors thank Dr. P. Chang and Dr. B. Kirtman for the model data and the constructive discussion during the visit at Texas A&M University.

References

Beena, B. S., and von Storch, J. S., 2009. Effects of fluctuating daily surface fluxes on the time-mean oceanic circulation.Climate Dynamics, 33 (1): 1-18, DOI: 10.1007/S00382-009-0575-Y.

Bengtsson, L., Hodges, K. I., and Keenlyside, N., 2010. Will extratropical storms intensify in a warmer climate? Journal of Climate, 22 (9): 2276-2301, DOI: 10.1175/2008jcli2678.1.

Biastoch, A., Boning, C. W., Getzlaff, J., Molines, J. M., and Madec, G., 2008. Causes of interannual-decadal variability in the meridional overturning circulation of the midlatitude North Atlantic Ocean. Journal of Climate, 21 (24): 6599-6615, DOI:10.1175/2008jcli2404.1.

Chang, E. K. M., 2007. Assessing the increasing trend in Northern Hemisphere winter storm track activity using surface ship observations and a statistical storm track model. Journal of Climate, 20 (22): 5607-5628, DOI: 10.1175/2007jcli1596.1.

Cunningham, S. A., Kanzow, T., Rayner, D., Baringer, M. O.,Johns, W. E., Marotzke, J., Longworth, H. R., Grant, E. M.,Hirschi, J. J.-M., Beal, L. M., Meinen, C. S., and Bryden, H.L., 2007. Temporal variability of the Atlantic meridional overturning circulation at 26.5˚N. Science, 317 (5840): 935-938.

Danabasoglu, G., 2008. On multidecadal variability of the Atlantic meridional overturning circulation in the Community Climate System Model Version 3. Journal of Climate, 21 (21):5524-5544, DOI: 10.1175/2008jcli2019.1.

Delworth, T. L., and Greatbatch, R. J., 2000. Multidecadal thermohaline circulation variability driven by atmospheric surface flux forcing. Journal of Climate, 13 (9): 1481-1495.

Kirtman, B. P., Straus, D. M., Min, D., Schneider, E. K., and Siqueira, L., 2009. Toward linking weather and climate in the interactive ensemble NCAR climate model. Geophysical Research Letters, 36: L13705, DOI: 10.1029/2009GL038389.

Knight, J. R., Allan, R. J., Folland, C. K., Vellinga, M., and Mann,M. E., 2005. A signature of persistent natural thermohaline circulation cycles in observed climate. Geophysical Research Letters, 32 (20): 242-257, DOI: 10.1029/2005gl024233.

Kwon, Y. O., and Frankignoul, C., 2012. Stochastically-driven multidecadal variability of the Atlantic meridional overturning circulation in CCSM3. Climate Dynamics, 38 (5-6): 859-876, DOI: 10.1007/S00382-011-1040-2.

Landsea, C. W., Pielke, R. A., Mestas-Nunez, A., and Knaff, J.A., 1999. Atlantic basin hurricanes: Indices of climatic changes.Climatic Change, 42 (1): 89-129.

Lozier, M. S., 2010. Deconstructing the conveyor belt. Science,328 (5985): 1507-1511, DOI: 10.1126/Science.1189250.

McManus, J. F., Francois, R., Gherardi, J. M., Keigwin, L. D.,and Brown-Leger, S., 2004. Collapse and rapid resumption of Atlantic meridional circulation linked to deglacial climate changes. Nature, 428 (6985): 834-837, DOI: 10.1038/Nature 02494.

Men, G., Wan, X., and Liu, Z., 2016. Tropical Atlantic climate response to different freshwater input in high latitudes with an ocean-only general circulation model. Journal of Ocean University of China, 15 (5): 751-757.

Mu, L., Chen, X. E., Song, J., Huan, L. I., Yan, L. I., and Jiang,X. Y., 2011. Study on mechanism of interdecadal Atlantic thermohaline circulation variability III: Influences of interdecadal variations by ocean-atmosphere elements. Acta Oceanologica Sinica, 33 (4): 19-28 (in Chinese with English abstract).

Pickart, R. S., Spall, M. A., Ribergaard, M. H., Moore, G. W. K.,and Milliff, R. F., 2003. Deep convection in the Irminger Sea forced by the Greenland tip jet. Nature, 424 (6945): 152-156,DOI: 10.1038/Nature01729.

Schmittner, A., Latif, M., and Schneider, B., 2005. Model projections of the North Atlantic thermohaline circulation for the 21st century assessed by observations. Geophysical Research Letters, 32 (136): 170-177.

Shaman, J., Samelson, R. M., and Skyllingstad, E., 2010. Airsea fluxes over the Gulf Stream region: Atmospheric controls and trends. Journal of Climate, 23 (23): 2651-2670, DOI: 10.1175/2010jcli3269.1.

Sriver, R. L., and Huber, M., 2007. Observational evidence for an ocean heat pump induced by tropical cyclones. Nature,447 (7144): 577-580, DOI: 10.1038/Nature05785.

Srokosz, M. A., and Bryden, H. L., 2015. Observing the Atlantic meridional overturning circulation yields a decade of inevitable surprises. Science, 348 (6241): 1255575.

Sutton, R. T., and Hodson, D. L. R., 2005. Atlantic Ocean forcing of North American and European summer climate. Science, 309 (5731): 115-118, DOI: 10.1126/Science.1109496.

Våge, K., Pickart, R. S., Thierry, V., Reverdin, G., Lee, C. M.,Petrie, B., Agnew, T. A., Wong, A., and Ribergaard, M. H.,2009. Surprising return of deep convection to the subpolar North Atlantic Ocean in winter 2007–2008. Nature Geoscience, 2 (1): 67-72.

Wan, X., Chang, P., Saravanan, R., Zhang, R., and Schmidt, M.W., 2009. On the interpretation of Caribbean paleo-temperature reconstructions during the Younger Dryas. Geophysical Research Letters, 36 (2): L02701, DOI: 10.1029/2008GL035 805.

Wen, C., Chang, P., and Saravanan, R., 2011. Effect of Atlantic meridional overturning circulation on tropical Atlantic variability: A regional coupled model study. Journal of Climate,24 (13): 3323-3343.

Woollings, T., Gregory, J. M., Pinto, J. G., Reyers, M., and Brayshaw, D. J., 2012. Response of the North Atlantic storm track to climate change shaped by ocean-atmosphere coupling. Nature Geoscience, 5 (5): 313-317, DOI: 10.1038/Ngeo1438.

Wunsch, C., 2002. What is the thermohaline circulation? Science, 298 (5596): 1179-1181.

Yin, J. H., 2005. A consistent poleward shift of the storm tracks in simulations of 21st century climate. Geophysical Research Letters, 32 (18): 109-127, DOI: 10.1029/2005gl023684.