1. Introduction
Coupled climate models are commonly used to make predictions of changes in the climate system. In the third assessment report of Houghton et al. (2001), coupled models began to be developed without flux adjustment and with a reasonably stable climate (e.g., Gordon et al. 2000; Boville and Gent 1998). With the use of flux adjustment techniques, ocean water masses are less likely to drift away from climatological values as the surface properties are constrained to be close to observed values. It is therefore important to quantify the response when flux adjustments are not used, as the drifts have the potential to be significantly larger. The choice of initial conditions and their imbalance with the atmospheric state is also key to the adjustment of a climate model toward equilibrium.
Coupled climate models can be initialized in a preindustrial state in a variety of ways (Stouffer et al. 2004). They are commonly integrated with ocean initial conditions either prescribed by climatological temperature and salinity distributions or from an ocean model partly spun up from these distributions forced by constant surface fluxes. The Hadley Centre Global Environmental Model version 1 (HadGEM1; Johns et al. 2006) has been developed to run without the use of flux adjustment. The approach taken in producing a preindustrial state has been very direct; the ocean model component has been initialized at rest (i.e., the three-dimensional velocity field is set to zero) with temperature and salinity fields from Levitus et al. (1998). The initial ocean state has been coupled on a daily time scale to an atmosphere model component with preindustrial forcing. The ocean velocity field typically adjusts to the density field in less than a decade (Johns et al. 2006).
In the Third Hadley Centre Coupled Ocean–Atmosphere GCM (HadCM3; Gordon et al. 2000), it was found that the implied and actual freshwater fluxes came into balance on a time scale of approximately 350 yr (Pardaens et al. 2003). Excessive salt accumulated in the tropical Atlantic Ocean and was transported to high latitudes where it was then convected downward into the deep ocean and transported away. The adjustment of the salinity field allowed the ocean transport of freshwater to balance the surface flux of freshwater.
Here we examine the adjustment of HadGEM1 toward an equilibrium state. This study is a follow-up to Johns et al. (2006) and examines the adjustment of ocean water masses and the impact on the radiative equilibrium in more detail. We look at the first 600 yr of the HadGEM1 integration. It should be noted that on time scales longer than 600 yr, the deep ocean water masses will continue to adjust and may have a significant impact on sea level. The HadGEM1 model is described in more detail in section 2. In section 3 we describe the adjustment of the radiative balance, and in section 4 we describe the oceanic mechanisms, which lead to adjustments in the North Atlantic. In section 5 we summarize the results and discuss the implications of these results for control experiments of coupled climate models.
2. Coupled climate model HadGEM1
HadGEM1 (Johns et al. 2006) is a coupled climate model run without the use of flux adjustment. HadGEM1 is based on the Hadley Centre Global Atmospheric Model (HadGAM1; Martin et al. 2006) coupled to a z-coordinate ocean model with 40 vertical levels and a horizontal resolution of 1° in longitude and 1° decreasing to 1/3° at the equator in latitude. The ocean model has fourth-order advection and diffusion numerics. The model includes the Gent et al. (1995) parameterization of oceanic eddies, which is believed to be partly responsible for allowing the ocean to transport sufficient heat to balance the surface heat fluxes given by the atmosphere (Gordon et al. 2000). The ocean model has a linear free surface (Dukowicz and Smith 1994) and surface freshwater fluxes are imposed as a change in the free surface height rather than as salt fluxes [which can be shown to be conservative over annual time scales; Roullet and Madec (2000)]. The sea ice in HadGEM1, which is described in McLaren et al. (2006), includes an ice thickness distribution (Lipscomb 2001) and dynamics based on the elastic–viscous–plastic rheology (Hunke and Dukowicz 1997).
The experiment discussed here is the HadGEM1 control run with fixed 1860 (preindustrial) forcing levels for greenhouse gases, ozone, sulphur, and other aerosol precursor emissions and land surface boundary conditions. The atmosphere was initialized from an analyzed state corresponding to 1 September 1998. The ocean temperature and salinity were initialized from climatological September values (Levitus et al. 1998) and sea ice was initialized from the Hadley Centre Sea Ice and SST (HadISST; Rayner et al. 2003) dataset in the Northern Hemisphere and HadCM3 in the Southern Hemisphere. Both ocean and atmosphere were initialized from rest. An error in the coupling of river outflow at some grid boxes leads to a downward drift in free surface height, which does not affect the simulation (Johns et al. 2006).
Figure 1a shows the top-of-the-atmosphere (TOA) radiation for the HadGEM1 control run. The TOA flux is initially negative but quickly adjusts toward a positive flux, which then transitions from 0.31 W m−2 in years 0–200, 0.20 W m−2 in years 200–400, and reaches a quasi equilibrium of 0.12 W m−2 in years 400–600. This leads to a long-term warming of the climate system. Since the heat capacity of the atmosphere is small this implies that the ocean will gain heat. Over the first 350 yr, the ocean heat content increase is 1.05 × 1024 J, which is equivalent to a surface heat flux of 0.27 W m−2.
3. Adjustment of TOA in HadGEM1
Throughout the experiment SWD remains constant at 341.4 W m−2. LWU_clear decreases by almost 3 W m−2 in the first 50 yr before increasing by 0.7 W m−2 over the next 500 yr. To compensate, SWU_clear (Table 1; Fig. 1b) increases over the first 50 yr by more than 2 W m−2 and then decreases by 0.3 W m−2 over the next 500 yr. CRF shows a continual increase of 0.1 W m−2 in the first 50 yr and another 0.1 W m−2 over the subsequent 500 yr. It is a delicate sum of these large terms which leads to the adjustment of the TOA toward zero.
a. Outgoing longwave radiation and sea surface temperature
Outgoing longwave radiation (LWU_clear) will increase/decrease as the surface temperature (SST) increases/decreases. In the initial 50 yr, global mean SST cools by 0.8°C (Fig. 2a). Applying Stefan’s law, and scaling for the ocean covering two-thirds of the Earth’s surface, this could lead to a decrease in outgoing longwave radiation of 2.9 W m−2 over the ocean, in agreement with the modeled change in LWU_clear. The reasons for the initial SST cooling are complex; it is partly due to initially applying a preindustrial atmospheric forcing to a present-day ocean state (this is estimated to account for 0.16°C cooling) but also due to excessively strong atmospheric wind forcing (Johns et al. 2006). The magnitude of vertical mixing in the ocean mixed layer is also a topic of ongoing research (M. Joshi 2006, personal communication).
Between years 200 and 350, global mean SST increases by approximately 0.2°C (Fig. 2a), which is equivalent to an increase of 0.7 W m−2 in LWU_clear, in agreement with the modeled change. The Northern and Southern Hemisphere contributions to the global mean SST show that the Northern Hemisphere is dominating the increase in SST (Fig. 3), in contrast with increases in heat content, which reach a maximum at 40°S. The Northern Hemisphere is responsible for 56% of the global SST change even though it covers only 42% of the ocean surface area. The SST increase in the Northern Hemisphere is largely confined to the North Atlantic subpolar gyre and the Greenland Sea (Fig. 4). The location of the SST changes suggests that it is related to the northernmost part of the thermohaline circulation (THC). The link between SST change and the THC will be discussed further in section 4.
From the time series of the Northern and Southern Hemisphere SSTs (Fig. 2b) we observe that the difference in the hemispheric temperatures indicates when the THC (Fig. 5a) is in a weak or strong state. The THC is defined as the maximum overturning between 20° and 30°N above 3000 m. For example, at year 250, the Northern Hemisphere is relatively cool and the Southern Hemisphere is relatively warm, associated with a weak THC (see Fig. 5a), while at year 400, the Northern Hemisphere is relatively warm and the Southern Hemisphere is relatively cool, associated with a strong THC. The correlation between the 20-yr mean hemispheric temperature difference (Fig. 2b) and the 20-yr mean THC (Fig. 5a) in HadGEM1 is 0.85. This high correlation has also been seen in other models (e.g., Latif et al. 2004; Vellinga and Wu 2004).
b. Shortwave radiation and sea ice
The outgoing shortwave radiation (SWU_clear) responds to the surface albedo of which sea ice is a dominant factor. Global sea ice area increases rapidly in the first 50 yr of the integration from 2 to 2.5 × 1013 m2 reaching a maximum of 2.6 × 1013 m2 after 20 yr (Fig. 2c). However, as surface temperature warms between years 200 and 350, the sea ice retreats (Fig. 4) with the total ice area reducing to 2.35 × 1013 m2. The largest reductions in sea ice area occur in regions of relatively low ice concentration and where the SST increases are greatest (Fig. 4).
The ice-albedo feedback process means that a reduction (increase) in sea ice area will lead to a reduction (increase) in the surface albedo, which should subsequently reduce (increase) SWU_clear. The radiative warming (cooling) due to the shortwave changes is usually partly offset by radiative cooling (warming) due to longwave changes (Fig. 1b) as also seen in Vellinga et al. (2002).
By calculating the change in surface outgoing shortwave fluxes over ice-covered and ice-free areas, we estimate the relative contributions of changes in ice and snow coverage to SWU_clear. In the initial 50 yr, SWU_clear increases by 2.2 W m−2. An increase in ice area of 0.55x1013 m2 is responsible for 80% of the change in SWU_clear, while changes in snow cover over land are responsible for 15% of the change in SWU_clear. Between years 200 and 350, the change in ice area is responsible for 65% of the change in SWU_clear, while 22% of the remaining change in SWU_clear can be attributed to changes in snow cover over land.
The Arctic ice volume varies over the first 400 yr of the HadGEM1 simulation (Fig. 5b). The anomalies in ice thickness are mainly confined to the Arctic basin and along the east Greenland coast. Comparison with the THC (Fig. 5a) shows that the ice volume and the THC are strongly anticorrelated; i.e., when the THC is strong (advecting warm water northward into the Arctic), sea ice volume is reduced and when the THC is weak (with reduced advection of warm water into the Arctic), sea ice volume increases. A similar anticorrelation was found by Jungclaus et al. (2005), over several centuries of a coupled model integration. The correlation between the THC and Arctic ice volume is at a maximum when the THC leads the Arctic ice volume by 9 yr.
4. Oceanic mechanisms leading to radiative balance
We have seen that HadGEM1 has a net positive TOA, which adjusts to less than 0.15 W m−2 on a 350-yr time scale. We have shown that this adjustment is linked to both longwave radiation and SST warming in the subpolar North Atlantic and shortwave radiation with a reduction in sea ice cover over the Arctic. The multicentury time scale suggests that the ocean plays a dominant role in this adjustment. In this section we will describe the role played by the ocean.
As discussed previously, in HadGEM1 the climate system is gaining heat overall. Since the heat capacity of the atmosphere is relatively low compared to that of the ocean, this implies that most of the residual heat flux at the top of the atmosphere will enter the ocean. We have examined the surface heat flux in HadGEM1 in three latitude bands: southern oceans (90°–30°S), tropics (30°S–30°N), and northern oceans (30°–90°N). Table 2 shows that, compared with da Silva et al. (1994) observations, the net heat gain in HadGEM1 is driven primarily by excess heat gain in the tropics and northern oceans. However, given limited flux measurements in the Southern Ocean and the imbalance in the climatology, it is likely that there is also net heat gain (relative to da Silva et al.) in the Southern Ocean.
The drifts in HadGEM1 in temperature and salinity in each latitude band (Fig. 6) show that, while the top few hundred meters of the ocean become cooler and fresher, the subsurface ocean becomes warmer and saltier. This shows that the excess heat and salt gain by the ocean is accumulating in the subsurface. There is a strong correlation between warming and salting, which suggests that the temperature and salinity drifts are largely density compensating. We confine our view to the top 2000 m as at greater depths the equilibrium time scale is O(1000 yr), which is significantly longer than the 600 yr discussed in this paper. North of 30°S, we also focus our attention on the Atlantic basin as this is where the largest changes are seen. As temperature and salinity changes are well-correlated we will often examine salinity changes as they are more easily identified as a tracer.
a. Southern Ocean
In the Southern Ocean (Figs. 6a,b) the subsurface warm, salty anomaly is in the depth range of Antarctic Intermediate Water (AAIW). The erosion of AAIW is first seen as a saline anomaly in the fifth month of the experiment at 37°S, 15°E. This anomaly grows rapidly and spreads, filling the subtropical gyre and eroding the Atlantic salinity minimum. Subsequent salting of the intermediate depth Atlantic is rapid; Fig. 7 shows that by year 100 the salinity minimum has retreated to 28°S. Examination of the salinity field along 37°S south of Africa (Fig. 8) reveals that by year 10 much of the salinity minimum has eroded, although there is little change in the density field.
The salinity increase in the AAIW layer in the Atlantic reflects salt being advected from the east. The salinity increase cannot be attributed to atmospheric forcing as the surface freshwater flux does not change and the maximum wind stress is located at approximately the same latitude as observations suggest (Russell et al. 2006). The zonal velocity field at 37°S reveals a strong westward flow from the Indian Ocean to the Atlantic extending from 25°E to 5°W. In agreement with observations the model Indian Ocean is warmer and has more saline than the Atlantic; therefore, this flow transports salt and heat into the South Atlantic. In year 1 of the simulation, the salinity difference between the two basins, in the Agulhas region, is largest in the AAIW layer with a difference of 0.14 psu on the 27.2 isopycnal, the core of the model AAIW. This advection of salt by the mean flow results in the volume-averaged salinity in the AAIW layer, between 35° and 40°S, increasing at a rate of 0.02 psu yr−1. By year 50 the Atlantic at these latitudes is saltier and warmer than in the Indian basin on all isopycnals denser than 26.3. After 350 yr the difference in salinity is still largest in the AAIW layer, but the Atlantic is 0.34 psu fresher than the Indian Ocean. During the first 10 yr of the simulation the salinity on AAIW isopycnals in the southwest Indian Ocean has not drifted significantly; by year 10 they are slightly fresher than the climatology. This implies that the erosion in the Atlantic is driven by the mean flow rather than a salinity error in the Indian Ocean.
The flow off South Africa is dominated by the Agulhas Current and its retroflection. The Agulhas Current flows westward along the southern coast of Africa with the majority of the flow retroflecting back into the Indian Ocean as the Agulhas Return Current. The strength of the Agulhas Current is well simulated by HadGEM1. Bryden et al. (2005) cite a mean transport of 69.7 Sv (1 Sv ≡ 106 m3 s−1) at 31°S in the Indian Ocean and at the same latitude the mean flow in the first decade of HadGEM1 is 68.8 Sv. Not all of the water carried in the Agulhas Current retroflects; the part which continues into the Atlantic is termed the Agulhas Leakage. This flux is poorly constrained by observations, with much of the transport into the Atlantic occurring within large anticyclonic eddies (Gordon et al. 1992). The volume exchange between the two oceans is thought to be around 10% of the Agulhas transport (Dijkstra and de Ruijter 2001). In HadGEM1 there is little retroflection with the majority of the flow continuing westward into the Atlantic basin (Fig. 9). At 25°E the Agulhas transport in the model is 50.4 Sv, 40% of which occurs in the AAIW layer (27.0–27.5 isopycnals). At 20°E where retroflection should occur (Dijkstra and de Ruijter 2001), the model exhibits a westward transport of 34.0 Sv.
In conclusion, the lack of a realistic Agulhas retroflection in HadGEM1 results in a strong westward transport of saline Indian Ocean water into the Atlantic Ocean, eroding the salinity minimum. The erosion is further enhanced by a spurious upwelling at 40°S, which is driven by topography on the Mid-Atlantic Ridge.
b. Tropical Atlantic
As AAIW erodes in HadGEM1 the tropical Atlantic fills with warmer, saltier water (Figs. 6c,d). The increase in salinity resulting from the erosion of AAIW is further enhanced by strong upwelling in the western equatorial Atlantic. The monthly wind stress in the tropical Atlantic is consistently overestimated compared with the Southampton Oceanography Centre (SOC) climatology (Josey et al. 1996). The error is largest when the intertropical convergence zone (ITCZ) shifts north in September and October and there is a strong intensification of the easterlies in the west of the basin. This results in vigorous upwelling and a salinity anomaly centered at 100 m in the western equatorial Atlantic. This anomaly is transmitted across the basin both on the rapid time scale of a Kelvin wave and with the mean flow. Strong diapycnal mixing then leads to a buildup of salinity in the subsurface.
Two processes in HadGEM1 act to further increase the salinity in both the surface and thermocline. First, ventilation causes the salinity to increase as the isopycnals outcrop in a region where HadGEM1 exhibits excess freshwater loss. The isopycnals outcrop in the western Atlantic between April and June 5°–10°N and 40°–50°W. There is strong evaporation in the region at this time. Second, the freshwater outflow from the Amazon River is rapidly transported northward by the western boundary current in HadGEM1, causing a fresh error in the gyre and a saline error on the equator where the river outflows (Fig. 10). The shelf where the Amazon enters the model Atlantic is 50 m deep. Once on the shelf, the river outflow is subjected to the western boundary current and is not able to mix down. More realistic topography and a shelf model may be needed to mix the Amazon water realistically before it encounters the western boundary current.
c. North Atlantic
For the region north of 30°N, we focus on the North Atlantic as this is the basin that shows the most dramatic surface changes. The North Atlantic (Figs. 6e,f) shows warm anomalies centered at both 400 and 1400 m. The temperature and salinity changes at 400 m (Fig. 11a; salinity is not shown but is well correlated with temperature) show extensive warming (and salting) throughout the northern branch of the subtropical gyre. Although the North Atlantic Current has cooled (due to a southward shift in the edge of the subpolar gyre), warm anomalies can be seen in the Norwegian Current, around the Greenland Sea and into the northern part of the subpolar gyre. Warming at this depth exceeds 3°C in some areas. The 27.2 isopycnal outcrops at the surface in both the Southern Hemisphere at approximately 50°S and in the Greenland Sea. The 27.2 isopycnal is at a depth of around 400 m in the rest of the North Atlantic. Warm, salty anomalies from the South Atlantic and the tropics are advected into the North Atlantic and eventually outcrop in the region of the Greenland Sea. The outcrop region is collocated with where the warm SSTs (and high surface salinities) are observed (Fig. 4).
In terms of air–sea exchange, when the warm, salty water outcrops, a warm ocean surface and cold atmosphere leads to large surface heat loss in excess of 40 W m−2 over parts of the Greenland Sea and Labrador Sea (Fig. 11c). The surface heat loss acts to cool the surface and reduce the SST anomaly. The sea surface in the Greenland Sea also becomes much saltier, which is in contrast to the rest of the globe where freshening continues to dominate. The net change in freshwater fluxes over the North Atlantic region is a 65% decrease in net precipitation. In particular, over the central Labrador Sea there is a large decrease in net precipitation (Fig. 11d). The decrease in net precipitation in the Labrador Sea is collocated with the increase in surface heat loss and will act to enhance the surface salinity in the region though it will be diluted slightly due to an increase in precipitation and sea ice melt on the eastern edge of Greenland. Overall, the air–sea exchange leads to colder, saltier, and therefore denser surface water, which is prone to convection. In the Labrador Sea, the strength of convection is linked to doming of the isopycnals. Convection at year 350 is clearly stronger than at year 150; the 27.8 isopycnal is more strongly domed and the 27.9 isopycnal is higher in the water column, suggesting that the volume of North Atlantic Deep Water (NADW) has increased (Fig. 12).
Anomalies at 1400 m lag those at 400 m (Fig. 6e), with the maximum correlation over the first 350 yr between the two depths occurring when the drifts at 1400 m lag those at 400 m by 38 yr. The delay in appearance of anomalies at 1400 m is due to the formation of NADW (as described in the previous paragraph) and its subsequent southward flow throughout the North Atlantic (Fig. 11b). The largest change in the thermohaline circulation occurs between years 250 and 350 (Fig. 5a) when its strength increases from the minimum value in the time series of 13.7 Sv to the maximum value in the time series of 20.1 Sv.
As the intermediate water in the Atlantic becomes saltier, the salinity contrast between the Atlantic and Indian intermediate waters reverses and the Atlantic intermediate water temperature and salinity properties approach equilibrium. This means that once the anomalous water properties have emerged at the surface in the North Atlantic and both radiative and surface fluxes have adjusted, the thermohaline circulation can approach equilibrium. The poor Agulhas simulation drives the adjustment, acting as a mechanism to allow the system to reach a quasi equilibrium. The thermohaline circulation in the North Atlantic simply responds to the adjustment process.
Although the large increase in the THC between years 250 and 350 (Fig. 5a) is highly correlated with SST increase and sea ice reduction (Figs. 2a,c), there is also multidecadal variability in the THC time series. In HadCM3 the THC variability is related to the position of the ITCZ (Vellinga and Wu 2004). In HadGEM1, the equatorial Atlantic freshwater flux (as a proxy for the ITCZ) and the THC covary as expected (Figs. 5a,c) with a time scale of approximately 100 yr. When the ITCZ is northward, the freshwater flux into the tropical equatorial Atlantic is large, and while the tropics freshen, the THC strengthens. The fresh anomaly is eventually advected northward and the THC weakens. As the ITCZ goes into the reverse phase, the tropics become saltier, eventually leading to a strengthening of the THC.
d. Summary
The above results demonstrate that in HadGEM1 the major adjustment of the ocean water masses is driven by the lack of an Agulhas retroflection. The lack of a retroflection leads to an anomalous input of warm, salty water from the Indian Ocean into the Atlantic Ocean. As seen in ocean-only experiments by Weijer et al. (2002), the anomalous water mass source in the South Atlantic is quickly advected northward throughout the Atlantic. In fact, HadGEM1 shows a similar pattern of advection to that shown in Fig. 10 of Weijer et al. (2002), but the time scales in HadGEM1 are longer. When the warm, salty water outcrops it increases SST, leads to reduction of Arctic sea ice and, through changes in convection, the magnitude of the overturning in the Atlantic is modified.
5. Discussion
In this paper we have described the adjustment of the coupled model HadGEM1 toward radiative equilibrium. HadGEM1 was initialized with climatological conditions for ocean temperature and salinity and the ocean and atmosphere were coupled from rest with daily coupling. HadGEM1 has a positive (incoming) flux at the top-of-the-atmosphere, which adjusts toward zero over a 350-yr-plus period of the control experiment. In this paper we demonstrate how the adjustment process is linked to the anomalous inflow of warm, salty intermediate water from the Indian and Southern Oceans. Antarctic Intermediate Water (or the warm, salty water replacing it) subsequently outcrops in the North Atlantic leading to increased SSTs. The SST changes lead to global adjustment of the radiative balance as well as a thermohaline circulation event. The events described in this paper bear some similarity to the thermohaline flushes described by Winton and Sarachik (1993). Winton and Sarachik (1993) described deep decoupled phases followed by thermohaline flushes and finally a long coupled phase. HadGEM1 does not have a decoupled phase, but the strengthening of the thermohaline circulation is very similar to a thermohaline flush.
a. Comparison with HadCM3
We can also compare what happens in HadGEM1 with HadCM3 (Gordon et al. 2000). HadCM3 has a negative (outgoing) flux at the top-of-the-atmosphere (Fig. 1c). HadCM3 clearly does not have an adjustment associated with the radiative balance and we can ask why this is the case. One explanation is that net heat loss by the ocean is likely to be more easily sustained than net heat gain; in the case of net heat loss, cold water is likely to sink into the deep ocean, which adjusts over multimillennial time scales. Net heat gain, as in HadGEM1, is more likely to be trapped in the upper parts of the ocean and subject to interaction with the atmosphere at submillennial time scales. Stouffer (2004) also finds that when subjected to cooling the Geophysical Fluid Dynamics Laboratory (GFDL) coupled model adjusted almost twice as fast as when subjected to warming. This agrees with the comparison of adjustment time scales in HadCM3 and HadGEM1; HadCM3 had a positive (downward) TOA flux in the first year but adjusted toward a quasi equilibrium in less than 200 yr (Fig. 1c), while HadGEM1 did not reach a quasi equilbrium until after 350 yr (Fig. 1a).
HadCM3 did, however, exhibit a 350-yr adjustment period in terms of the freshwater balance (Pardaens et al. 2003). In this case, the salinity distributions of the Atlantic adjusted to enable the freshwater transported by the ocean to balance the surface freshwater fluxes; the fresh tongue of AAIW extended into shallower depths than the initial climatology while NADW became saltier. Pardaens et al. (2003) attribute the high latitude salting in HadCM3 to excessive evaporation in the tropics, transporting salty water into the North Atlantic, which subsequently convects and forms salty North Atlantic Deep Water.
We have seen in HadGEM1 that the poor simulation of the Agulhas retroflection region plays a key role in leading the adjustment of the climate system. The velocity field in this region in HadCM3 looks very similar to that in HadGEM1, it lacks a retroflection and exhibits an overly zonal flow westward into the Atlantic. This flow results in erosion of the Antarctic Intermediate Water salinity minimum. However, in contrast to HadGEM1 a new salinity minimum emerges after 70 yr. The new intermediate water ventilates in the southwest Atlantic Ocean and reflects drifts to fresher and lighter conditions in the Southern Ocean. As HadCM3 adjusts, the Indian Ocean quickly becomes fresher than the Atlantic particularly at mode and intermediate water densities, further enhancing the new salinity minimum (S. Stark and I. Culverwell 2007, unpublished manuscript). In HadCM3, the zonal flow into the Atlantic moves from eroding to reinforcing Antarctic Intermediate Water.
b. Future development of climate models
What can we learn from these results for future development of climate models? A number of issues can be highlighted. First, we can ask whether climate models should be initialized from rest with present-day ocean temperature and salinity and preindustrial atmosphere conditions and then integrated for several centuries before a quasi equilibrium is reached? Transient runs are then started from the quasi equilibrium. This is the strategy that was adopted for HadCM3 and HadGEM1 but not necessarily for all coupled models. The problem with initializing from present-day ocean conditions is that the climate system is unlikely to be in balance and there is almost certainly going to be a climate drift. Stouffer et al. (2004) suggest initializing both atmosphere and ocean with present-day conditions and applying the observed radiative forcing backward in time until a preindustrial state is reached. This could be further improved by ensuring that the initial atmosphere and ocean conditions are close to being in balance with oceanic heat and freshwater transports being close to the transports implied by the atmosphere.
Second, we can ask, what are the priorities for model development in this family of models on the basis of these results? The simulation of the Agulhas retroflection needs to be improved as we have demonstrated that salt and heat transports from the Indian to the Atlantic Ocean impact both Atlantic water masses and the thermohaline circulation. The surface forcing provided by the atmosphere model needs to be improved, as the strong winds lead in part to excessive surface cooling. Improvements to the ocean vertical mixing may also be required as we cannot rule out that the surface cooling is in part due to vertical mixing.
Finally, we can ask, how long is an appropriate adjustment period of a coupled model before climate change experiments (possibly for the purpose of detection and attribution studies) are initialized? The problem with initializing climate change experiments from control model states before the water masses in the control experiment are close to equilibrium is that if nonlinear changes take place in the control experiment it increases the difficulty in distinguishing between model drift and a climate change signal. The results here, taken together with Pardaens et al. (2003), suggest that, at least for the Met Office climate models, an adjustment period on the order of 350 yr is required to allow oceanic adjustment to take place (Stouffer et al. 2004). This time scale will clearly be model dependent and also determined by the strategy adopted for initializing the ocean component.
c. Summary
Assuming that it is unlikely that a climate model will ever have a radiative balance exactly equal to zero, we should expect that the climate model will adjust on long time scales. The Atlantic is clearly an important region in terms of oceanic changes and it is therefore important to closely monitor intermediate waters as an early indicator of more widespread adjustments. On longer millennial time scales, adjustments in the deep ocean water masses [with ventilation time scales of O(1000 yr) in the North Pacific; England (1995)] will also come into play. In particular, adjustments in the deep ocean have the potential to impact sea level on long millennial time scales. Here we have assumed that the equilibrium state will be characterized by a radiation balance close to zero with small fluctuations. Haarsma et al. (2001) show that an intermediate complexity model can display low-frequency variations in the thermohaline circulation with a time scale of O(10 000) years. It is therefore plausible that a climate model could exhibit ultralow-frequency variability in the radiative balance.
This paper has described the adjustment of the HadGEM1 coupled climate model toward radiative equilibrium. The ocean plays the dominant role in the adjustment of the global heat balance by storing heat and subsequently releasing it to the atmosphere with a significant impact on the thermohaline circulation. In the case of HadGEM1, the lack of an Agulhas retroflection is a key model error, which leads to the adjustment of the thermohaline circulation and radiative balance. The long time scales involved in radiative equilibrium are important considerations when planning scenario experiments.
Acknowledgments
We thank Ron Stouffer and an anonymous reviewer for their insightful comments, Ian Culverwell for his contribution from sensitivity experiments, and Anne Pardaens, Gareth Jones, and Richard Wood for useful discussion. This work was supported by the U.K. Department for Environment, Food and Rural Affairs under contract PECD 7/12/37 and by the Government Meteorological Research and Development Programme.
REFERENCES
Boville, B. A., and P. R. Gent, 1998: The NCAR Climate System Model, version one. J. Climate, 11 , 1115–1130.
Bryden, H. L., L. M. Beal, and L. M. Duncan, 2005: Structure and transport of the Agulhas Current and its temporal variability. J. Oceanogr., 61 , 479–492.
da Silva, A. M., C. C. Young-Molling, and S. Levitus, 1994: Algorithms and Procedures. Vol. 1, Atlas of Surface Marine Data, NOAA Atlas NESDIS 6, 83 pp.
Dijkstra, H. A., and W. P. M. de Ruijter, 2001: On the physics of the Agulhas Current: Steady retroflection regimes. J. Phys. Oceanogr., 31 , 2971–2985.
Dukowicz, J. K., and R. D. Smith, 1994: Implicit free-surface method for the Bryan-Cox-Semtner ocean model. J. Geophys. Res., 99 , 7991–8014.
England, M. H., 1995: The age of water and ventilation timescales in a global ocean model. J. Phys. Oceanogr., 25 , 2756–2777.
Gent, P. R., J. Willebrand, T. J. McDougall, and J. C. McWilliams, 1995: Parameterizing eddy-induced tracer transports in ocean circulation models. J. Phys. Oceanogr., 25 , 463–474.
Gordon, A. L., R. F. Weiss, W. M. Smethie Jr., and M. J. Warner, 1992: Thermocline and intermediate water communication between the South Atlantic and Indian Oceans. J. Geophys. Res., 97 , 7223–7240.
Gordon, C., C. Cooper, C. A. Senior, H. Banks, J. M. Gregory, T. C. Johns, J. F. B. Mitchell, and R. A. Wood, 2000: The simulation of SST, sea ice extents and ocean heat transports in a version of the Hadley Centre coupled model without flux adjustments. Climate Dyn., 16 , 147–168.
Haarsma, R. J., J. D. Opsteegh, F. M. Selten, and X. Wang, 2001: Rapid transitions and ultra-low frequency behaviour in a 40kyr integration with a coupled climate model of intermediate complexity. Climate Dyn., 17 , 559–570.
Houghton, J. T., Y. Ding, D. J. Griggs, M. Noguer, P. J. van der Linden, X. Dai, K. Maskell, and C. A. Johnson, 2001: Climate Change 2001: The Scientific Basis. Cambridge University Press, 881 pp.
Hunke, E. C., and J. K. Dukowicz, 1997: An elastic–viscous–plastic model for sea ice dynamics. J. Phys. Oceanogr., 27 , 1849–1867.
Johns, T. C., and Coauthors, 2006: The new Hadley Centre climate model (HadGEM1): Evaluation of coupled simulations. J. Climate, 19 , 1327–1353.
Josey, S. A., E. C. Kent, D. Oakley, and P. K. Taylor, 1996: A new global air-sea heat and momentum climatology. International WOCE Newsletter, No. 24, WOCE International Project Office, Southampton, United Kingdom, 3–5.
Jungclaus, J. H., H. Haak, M. Latif, and U. Mikolajewicz, 2005: Arctic–North Atlantic interactions and multidacadal variability of the meridional overturning circulation. J. Climate, 18 , 4013–4031.
Latif, M., and Coauthors, 2004: Reconstructing, monitoring, and predicting multidecadal-scale changes in the North Atlantic thermohaline circulation with sea surface temperature. J. Climate, 17 , 1605–1614.
Levitus, S., and Coauthors, 1998: Introduction. Vol. 1, World Ocean Database 1998, NOAA Atlas NESDIS 18, 346 pp.
Lipscomb, W. H., 2001: Remapping the thickness distribution in sea ice models. J. Geophys. Res., 106 , 13989–14000.
Martin, G. M., M. A. Ringer, V. D. Pope, A. Jones, C. Dearden, and T. J. Hinton, 2006: The physical properties of the atmosphere in the new Hadley Centre Global Environmental Model (HadGEM1). Part 1: Model description and global climatology. J. Climate, 19 , 1274–1301.
McLaren, A. J., and Coauthors, 2006: Evaluation of the sea ice simulation in a new atmosphere-ocean coupled climate model (HadGEM1). J. Geophys. Res., 111 .C12014, doi:10. 1029/2005JC003033.
Pardaens, A. K., H. T. Banks, J. M. Gregory, and P. R. Rowntree, 2003: Freshwater transports in HadCM3. Climate Dyn., 21 , 177–195.
Rayner, N. A., D. E. Parker, E. B. Horton, C. K. Folland, L. V. Alexander, D. P. Rowell, E. C. Kent, and A. Kaplan, 2003: Global analyses of sea surface temperature, sea ice, and night marine air temperature since the late nineteenth century. J. Geophys. Res., 108 .4407, doi:10.1029/2002JD002670.
Roullet, G., and G. Madec, 2000: Salt conservation, free surface, and varying levels: A new formulation for ocean general circulation models. J. Geophys. Res., 105 , 23927–23942.
Russell, J. L., R. J. Stouffer, and K. W. Dixon, 2006: Intercomparison of the Southern Ocean circulation in IPCC coupled model control simulations. J. Climate, 19 , 4560–4575.
Stouffer, R. J., 2004: Time scales of climate response. J. Climate, 17 , 209–217.
Stouffer, R. J., A. J. Weaver, and M. Eby, 2004: A method for obtaining pre-twentieth century initial conditions for use in climate change studies. Climate Dyn., 23 , 327–339.
Vellinga, M., and P. Wu, 2004: Low-latitude freshwater influence on centennial variability of the Atlantic thermohaline circulation. J. Climate, 17 , 4498–4511.
Vellinga, M., R. A. Wood, and J. M. Gregory, 2002: Processes governing the recovery of a perturbed thermohaline circulation in HadCM3. J. Climate, 15 , 764–780.
Weijer, W., W. P. M. de Ruijter, A. Sterl, and S. S. Drijfhout, 2002: Response of the Atlantic overturning circulation to South Atlantic sources of buoyancy. Global Planet. Change, 34 , 293–311.
Winton, M., and E. S. Sarachik, 1993: Thermohaline oscillations induced by strong steady salinity forcing of ocean general circulation models. J. Phys. Oceanogr., 23 , 1389–1410.
Time series of TOA flux in HadGEM1 for (a) net and (b) anomaly of components (with respect to the 600-yr mean of each component) and (c) HadCM3 for 20-yr means. Prior to year 11, annual means are shown.
Citation: Journal of Climate 20, 23; 10.1175/2007JCLI1688.1
Time series of (a) global average sea surface temperature, (b) difference between the Northern and Southern Hemisphere sea surface temperature, and (c) total ice area for annual and 20-yr means in HadGEM1.
Citation: Journal of Climate 20, 23; 10.1175/2007JCLI1688.1
Latitudinal change in SST (°C) and heat content (1022 J degree−1) between the average of years 346–355 and 11–20. The first 10 yr are discarded because of the initial cooling.
Citation: Journal of Climate 20, 23; 10.1175/2007JCLI1688.1
Change in SST in the Northern Hemisphere between the average of years 346–355 and 11–20. The region where sea ice concentration has decreased by more than 10% is shown within the blue contour. The outcrop region for the 27.2 isopycnal (defined as where the isopycnal is less than 50 m deep in ice-free ocean) is shown within the black contour. Note nonlinear scale.
Citation: Journal of Climate 20, 23; 10.1175/2007JCLI1688.1
Time series of (a) meridional overturning, (b) Arctic ice volume, and (c) freshwater flux between 0° and 15°N in equatorial Atlantic for annual and 20-yr means in HadGEM1.
Citation: Journal of Climate 20, 23; 10.1175/2007JCLI1688.1
Temperature and salinity drifts as a function of depth and time for (a), (b) southern oceans (90°–30°S), (c), (d) tropical Atlantic (30°S–30°N), and (e), (f) North Atlantic (30°–90°N). For the tropical Atlantic, the scale is 4 times that shown.
Citation: Journal of Climate 20, 23; 10.1175/2007JCLI1688.1
Time series of the most northerly extent of AAIW at 20°W. AAIW is defined here as a salinity of 34.7 psu on the 27.2 isopycnal (the core of the model AAIW).
Citation: Journal of Climate 20, 23; 10.1175/2007JCLI1688.1
Salinity (colors) and density (contours) between 15°W and 30°E along 37°S for years 1, 5, and 10 of HadGEM1.
Citation: Journal of Climate 20, 23; 10.1175/2007JCLI1688.1
HadGEM1 velocity field on the 27.2 isopycnal (the core of the model AAIW) for years 1–10.
Citation: Journal of Climate 20, 23; 10.1175/2007JCLI1688.1
Atlantic sea surface salinity minus climatology between 15°S and 30°N in years 345–55 of HadGEM1 with velocity overlain. The fresh tongue resulting from the Amazon outflow being advected by the western boundary current is clearly visible.
Citation: Journal of Climate 20, 23; 10.1175/2007JCLI1688.1
Change in temperature (°C) between years 11–20 and 346–355 for (a) 399- and (b) 1460-m depth. The currents (cm s−1) at this level for years 11–20 are superimposed. Change in (c) surface heat flux (W m−2) and (d) freshwater flux (kg m−2 s−1) between the average of years 346–355 and 11–20.
Citation: Journal of Climate 20, 23; 10.1175/2007JCLI1688.1
Potential density cross sections at 60°N through the Labrador Sea for (a) year 1, (b) average of years 146–155, and (c) average of years 346–355.
Citation: Journal of Climate 20, 23; 10.1175/2007JCLI1688.1
Comparison of components of TOA flux (W m−2) for years 1, 41–60, and 401–600.
Comparison of heat (PW) fluxes from da Silva et al. (1994) climatology and HadGEM1 (years 206–225) for the global ocean, southern oceans (90°–30°S), tropics (30°S–30°N), and northern oceans (30°–90°N).
