The adjustment of the coupled ocean–atmosphere system to a fresh water anomaly in the North Atlantic Ocean is investigated using a coupled GCM. In response to the anomaly the Atlantic Thermohaline Circulation (THC) collapses, and the associated reduction in northward heat transport causes sea surface temperatures (SST) to fall over the North Atlantic, and to increase over the tropical South Atlantic. Atmospheric feedbacks are excited as soon as the SST anomalies reach sufficient magnitude. A key stage in the adjustment process occurs 4–6 years after the perturbation is introduced when a significant SST dipole develops in the tropical Atlantic. This dipole causes a southward shift of the ITCZ and leads, in year 7, to the triggering of an El Niño event. It is concluded that atmospheric feedbacks could spread globally the influence of a sudden change in the THC much more quickly and efficiently than could ocean processes alone.
 The Thermohaline Circulation (THC) is a key component of the climate system. It plays an important role in maintaining the mean climate by transporting large amounts of heat O((1015)) from low-to-high latitudes. Substantial changes in the THC and associated heat transport would provoke significant climate change [e.g., Manabe and Stouffer, 1995; Vellinga and Wood, 2002; Clark et al., 2002]. Anticipated future anthropogenic warming of the climate system has the potential to weaken the THC by reducing surface water density in the formation region of North Atlantic Deep Water (NADW), both through high latitude warming and through enhanced poleward moisture transport in the atmosphere [Houghton et al., 2001]. If the warming is strong enough and sustained long enough, a complete collapse of THC cannot be excluded [Manabe and Stouffer, 1993; Stocker and Schmittner, 1997]. In order to better assess the risks associated with possible future abrupt change it is vital to better understand the processes through which the climate system responds to a sudden change in the THC.
 The processes that govern the ocean response to a sudden change in the THC have been widely studied [Kawase, 1987; Doscher et al., 1994; Yang, 1999; Goodman, 2001; Johnson and Marshall, 2002]. The initial adjustment occurs via the propagation of Kelvin waves along the western boundary. Within a few months, the signal reaches the western equatorial Atlantic and then travels along the equator. Reaching the eastern boundary, the Kevlin waves propagate poleward along the coast into both hemispheres, generating Rossby waves that propagate westward and carry the signal into ocean interior. Because the Kelvin waves travel very quickly (2 ms−1) the influence of a sudden change in the North Atlantic can be felt in the coastal and equatorial regions of the Indian and Pacific Oceans within a decade or two, even though the time for an advective signal to reach outside of the Atlantic basin is an order of magnitude longer [e.g. Goodman, 2001]. This said, the magnitude of the fast remote response is probably small. The Kelvin waves are subject to damping as they propagate, and any response in the interior must await the (comparatively slow) propagation of Rossby waves from the eastern boundary. Based on results from an ocean GCM study, Goodman  suggests that, outside of the Atlantic basin, the fast response “probably could not be detected over … random noise”.
 While the ocean model studies have yielded considerable insights, an important limitation is that they neglect the role of the atmosphere. Understanding this role is important not only to establish the impacts on climate of a change in the THC but also because atmospheric feedbacks may play a vital role in the adjustment process itself. The ocean response to a THC change involves changes in SST [e.g. Manabe and Stouffer, 1995; Yang, 1999] that can influence the atmosphere. An important consequence is that the atmosphere may offer a bridge to propagate the influence of a THC change from the Atlantic basin to remote regions far more quickly and efficiently than processes in the ocean alone can achieve.
 This paper reports results from an experiment designed to elucidate the coupled ocean-atmosphere processes involved in the adjustment to a sudden change in the THC. We focus our analysis on the coupled processes that are excited in the first decade following the introduction of a substantial fresh water anomaly in the North Atlantic, and find - as anticipated - that atmospheric feedbacks play a critical role in the adjustment process.
2. Coupled Model and experiment
 The model used is a version of the UK Hadley Centre global coupled ocean atmosphere general circulation model HadCM3 [Gordon et al., 2000]. The resolution is 2.5° × 3.75° latitude–longitude with 19 vertical levels for the atmospheric component and 1.25° × 1.25° latitude–longitude with 20 levels for the oceanic component. The model does not require flux adjustments to maintain a stable climate and simulates many aspects of climate variability, including ENSO [Latif et al., 2001], with considerable realism.
 We analyse two experiments which differ only in the initial conditions in the ocean. For the control experiment the initial ocean conditions were taken from a long (multi-century) integration of the coupled model. For the perturbed experiment a salinity anomaly was introduced by freshening the upper 500 m of North Atlantic (55°N–85°N, 90°W–20°E) by 2 psu. This instantaneous salinity perturbation is equivalent to a freshwater pulse of about 3.0 × 1014 m3 (about 8 Sv*year, 1 Sv = 106 m3s−1). A similar experiment has been performed by Vellinga and Wood  with the same model, but they focused on the quasi-equilibrium response. Here we focus on the adjustment of the coupled system.
 In order to separate the THC response from internal variability we performed ensembles of 5 integrations for both experiments. Individual ensemble members share identical ocean initial conditions but there are small differences in the atmospheric initial conditions [Collins et al., 2002]. All simulations are run for 10 years. The response to the salinity perturbation is defined as the difference between the ensemble means of the perturbed and control experiments.
 Our main interest lies in the coupled interactions which - as we shall see - take several years to develop. First, however, we discuss briefly the adjustment processes that occur in the ocean over the first few months. The initial response to the salinity perturbation includes an anomalous anticyclonic circulation over the high latitudes of the North Atlantic, corresponding to a weakening of subpolar gyre in line with geostrophic balance. As expected, a boundary wave train propagates southward along the western boundary. Figure 1 shows that by day 15 the signal has reached the coast of Florida. By day 120, there are significant anomalies in the upper ocean currents in the equatorial Atlantic. Timeseries of the ocean currents (panels c and d) indicate that a significant anomaly in the western boundary current at Florida develops in less than a month, and a significant response in the equatorial Atlantic in 1–2 months.
 Accompanying the perturbation propagation, there is significant change in the meridional overturning circulation (MOC). The salinity perturbation caps the upper layers of the North Atlantic ocean with relatively low density water. This capping reduces convective mixing and the production of dense water in the sinking region of the THC. Thus the THC weakens almost instantaneously. A THC index, defined as the average MOC from 40°N–50°N over depth from 500 m to 1500 m, indicates a transport of 16–18 Sv in the control experiment. About a month after the salinity perturbation is applied this index has reached a value close to zero.
 We now consider the coupled adjustment processes that arise in response to the change in the THC. The reduction in the THC and associated ocean heat transport (not shown) leads to significant changes in ocean temperatures, and thereby changes in the atmosphere. Figure 2 shows the evolution of SST and sea level pressure (SLP) anomalies over the North Atlantic, and the tropical South Atlantic, over the first ten years. After introduction of the salinity perturbation, the North Atlantic becomes cooler, the cold anomalies reaching a maximum of about 2.0°C around year 5. These anomalies then start to decay, associated with recovery of the THC. Meanwhile, the tropical South Atlantic warms over the first 5–6 years, reaching a maximum anomaly of 0.5°C, before decaying quite rapidly. Figure 2 also shows that the changes in SST are accompanied by changes in SLP. In the first 5 years anomalous high pressure builds over the North Atlantic while low pressure anomalies develop over the tropical South Atlantic. The tropical SLP anomalies are in line with simple models [e.g., Lindzen and Nigam, 1987] of the atmospheric response to tropical SST anomalies.
Figure 3 shows the annual mean surface temperature and SLP anomalies for different years. In year 2, significant cooling is confined to the North Atlantic from 20°–50°N and the Norwegian Sea. The cooling over northern hemisphere land is generally smaller and not significant. Positive temperature anomalies over the northwest North Atlantic are associated with the reduction in convective mixing mentioned above. Panel e shows that, over the region of negative SST anomalies, positive anomalies in SLP have developed. As time progresses (years 2–4), the cooling over the North Atlantic ocean becomes stronger and more extensive as a result of the weakened THC and associated ocean heat transport. In addition, the cooling signal spreads downstream over the Eurasian continent, where temperature anomalies of 1–2°C develop. Accompanying these changes in surface temperature is the eastward spread of pressure anomalies. By year 4 significant high pressure anomalies appear over North Africa and Europe, and there is already an indication of a significant low pressure anomaly, with an associated SST anomaly, over the North Pacific.
 The warming of the tropical South Atlantic is already visible in year 4, and by year 6 a prominent North-South dipole of temperature anomalies can be seen in the tropical Atlantic. This pattern is interesting because it is known that the inter-tropical convergence zone (ITCZ) in the Atlantic is sensitive to variations in the cross-Equator SST gradient [e.g., Moura and Shukla, 1981]; thus we may anticipate significant effects on the tropical atmosphere in response to this dipolar anomaly. Figure 3g shows that the SST dipole is mirrored in a dipole of SLP anomalies. Lastly, in year 7 there is a further very interesting development. The dipolar pattern in the tropical Atlantic has disappeared, but prominent temperature anomalies have now appeared in the tropical Pacific ocean. The pattern of anomalies is typical of an El Niño event, suggesting - remarkably - that the perturbation we introduced into the THC has led to the triggering of El Niño. At this stage the climate impacts of the THC change are felt across virtually the whole globe.
 Further insight into the processes via which the THC-induced signal is spread through the tropical atmosphere is provided by Figure 4. This shows the response in precipitation. In year 6 the impact of the tropical Atlantic SST dipole on the Atlantic ITCZ can clearly be seen. The dipolar pattern of precipitation anomalies indicates that the ITCZ has shifted southwards. The precipitation anomalies over the tropical Pacific in year 6 are not statistically significant, but by year 7 a pattern of significant anomalies, characteristic of El Niño, is seen. How could the events in the tropical Atlantic in year 6 lead to the triggering of El Niño in year 7? Detailed investigation is beyond the scope of this paper, but a plausible mechanism is as follows. 11 The shift of the ITCZ in year 6 is associated with anomalies in the diabatic heating of the atmosphere. Figure 4a suggests that the peak of the heating anomalies will be off the Equator and in this situation simple theory [Gill, 1980] suggests that the atmospheric response is primarily governed by Rossby waves that will propagate westwards into the east Pacific. Here, they may induce anomalies in the near-equatorial surface windstress that could trigger El Niño. Westerly anomalies along the equator would be expected in response to the positive diabatic heating anomalies seen south of the Equator in the tropical Atlantic, so it may be that the Rossby wave forcing is dominated by this part of the dipolar pattern.
4. Discussion and Conclusions
 The aim of this investigation was to study the dynamic adjustment of the coupled ocean–atmosphere system to a sudden change (decrease) in the Atlantic THC. We have found that, in our experiment, the adjustment process involves four stages:
Years 0–2: Decrease of N. Atlantic SST and associated increase in SLP.
Years 2–4: Eastward spread of the North Atlantic changes over North Africa and Eurasia, leading to SLP anomalies over the North Pacific.
Years 4–6: Development of a dipolar temperature anomaly in the tropical Atlantic and associated southward shift of the ITCZ.
Year 7: El Niño event triggered in the tropical Pacific. Climate impacts are now global.
 A notable feature of these results is that our hypothesis - that atmospheric feedbacks could spread globally the influence of a change in the Atlantic THC much more quickly and efficiently than could ocean processes alone - is clearly supported. Furthermore, we have seen that the equatorial waveguide in the atmosphere may play a key role. If a change in the THC leads to significant changes in SST in the tropical Atlantic, then the equatorial waveguide offers an efficient means to propagate the signal throughout the global tropics and, from there, to the rest of the globe. It would be of considerable interest to investigate whether this mechanism could help to explain past changes in global climate.
 Also intriguing is the link we have found between the THC and El Niño. The significance of this result may not be limited to rapid climate change. If a sudden change in the THC can trigger El Niño then more gradual changes in the THC may also impact the tropical Pacific, perhaps modulating El Niño on multi-decadal timescales. In addition, there is the possibility of two-way interactions between the THC and El Niño. Latif  recently suggested that long-term changes in the tropical Pacific may influence the THC by modulating the transport of freshwater from the Atlantic to the Pacific Ocean. The detailed nature of THC-El Niño interactions, and their possible role in regulating climate, is an important area for future research.
 This study has been a first, relatively simple, investigation of the coupled processes involved in adjustment to a sudden change in the THC. It is likely that the details of the adjustment process, and especially the time required for each stage [see Doscher et al., 1994 regarding sensitivity to ocean model resolution], will be sensitive to the exact experimental design and the model used. We suspect, however, that the major features we have outlined could well be generic.
 This work was supported by the U. K. Universities Global Atmospheric Modelling Programme (UGAMP), the Royal Society and the EU PREDICATE project under contract EVK2-CT-1999-00020. We thank Matthew Collins for providing the control experiments.
An alternative to the hypothesis proposed here is that the changes in the North Pacific are critical to triggering El Niñ;o.