The role of the Indonesian Throughflow in the global climate system is investigated with a coupled ocean–atmosphere model by contrasting simulations with realistic throughflow and closed Indonesian passages.
The Indonesian Throughflow affects the oceanic circulation and thermocline depth around Australia and in the Indian Ocean as described in previous studies and explained by Sverdrup transports. An open throughflow thereby increases surface temperatures in the eastern Indian ocean, reduces temperatures in the equatorial Pacific, and shifts the warm pool and centers of deep convection in the atmosphere to the west. This control on sea surface temperature and deep convection affects atmospheric pressure in the entire Tropics and, via atmospheric teleconnections, in the midlatitudes. As a result, surface wind stress in the entire Tropics changes and meridional and zonal gradients of the tropical thermocline and associated currents increase in the Pacific and decrease in the Indian Ocean. The response includes an acceleration of the equatorial undercurrent in the Pacific, and a deceleration in the Indian Ocean. Thus the Indonesian Throughflow exerts significant control over the global climate in general and the tropical climate in particular.
Changes of surface fluxes in the Pacific warm pool region are consistent with the notion that shading by clouds, rather than increases of evaporation, limit highest surface temperatures in the open ocean of the western Pacific. In the marginal seas of the Pacific and in the Indian Ocean no such relationship is found. The feedback of the throughflow transport and its wind forcing is negative and suggests that this interplay cannot excite growing solution or lead to self-sustained oscillations of the ocean–atmosphere system.
The Indonesian Throughflow is the only low-latitude oceanic connection between the major ocean basins of today. Its transports of mass, heat, and freshwater from the Pacific into the Indian Ocean are important for the oceanic circulation and sea surface temperatures, but its role in the global climate system and its impact on simulations of the climate by coupled ocean–atmosphere models are unknown. In this study, the role of the Indonesian Throughflow in the global ocean–atmosphere system is examined by contrasting climates simulated by a coupled general circulation model with open and closed Indonesian passages.
The role of the throughflow in the ocean has been recently reviewed by Godfrey (1996), so only a brief summary is presented here. The mass transport of the throughflow affects the thermocline of the Indian Ocean and is in part responsible for anomalous deep thermocline and lack of cold upwelling off the western coast of Australia (Godfrey and Golding 1981). The heat transport of the throughflow removes a significant amount of the surface heat flux the Pacific received in the Tropics, one-third in a simulation of Hirst and Godfrey (1993) or 0.63 and 0.9 PW in the analysis of a coupled model by Schneider and Barnett (1997). The throughflow forces the Leeuwin current, either due to a particular form of vertical mixing (Kundu and McCreary 1986) or by atmospheric cooling and the ensuing geostrophic adjustment of warm waters supplied by the throughflow (Godfrey and Weaver 1991; Hughes et al. 1992). Salt transports by the throughflow represent a major salt sink for the Pacific and a gain for the Indian Ocean (Piola and Gordon 1984). Finally, water mass considerations (Gordon 1986; Gordon et al. 1992) and numerical experiments (Shriver and Hurlburt 1997) suggest that the throughflow takes part in the return branch of the thermohaline circulation.
These effects of the throughflow on the oceanic circulation were confirmed by Hirst and Godfrey (1993, hereafter referred to as HG), who compared simulations using ocean general circulation models with open and closed Indonesian passages. In addition, HG showed that the thermocline structures in the Indian Ocean and southern Atlantic were altered in a manner consistent with Sverdrup circulation (Godfrey 1989) and that surface temperatures changed in regions far removed from Indonesian waters where vertical mixing communicated changes of the thermocline to the surface. The signature of the Indonesian Throughflow on surface temperatures in the southeastern Indian Ocean and in the Great Australian Bight were confirmed by simulations of the Antarctic Ocean (Ribbe and Tomczak 1997).
In HG’s experiment, surface fluxes of heat and freshwater were approximated by Newtonian relaxation to observed values, and surface wind stresses were prescribed from observations. Thus changes of the surface fluxes of heat and freshwater were those obtained from this Newtonian formulation, and fluxes of momentum were held constant. The sensitivity of the coupled ocean–atmosphere system to the Indonesian Throughflow could therefore not be determined in HG’s study.
Results of HG, together with the observed atmospheric response to interannual anomalies of surface temperature, yield a hypothesis for the role of the throughflow in the global climate system. Observations of interannual anomalies show that an increase of surface temperature in the tropical central Pacific, a response seen by HG, causes an eastward shift of the centers of deep convection and indicate that the throughflow controls positions of the western Pacific warm pool and centers of deep convection. Atmospheric teleconnections between interannual anomalies of surface temperature in the equatorial Pacific and atmospheric pressure in midlatitudes (Horel and Wallace 1981; Karoly 1989) suggest that the throughflow affects midlatitudes by its control of tropical surface temperatures. Similarly, observed links between surface temperature in the tropical Pacific and the summer monsoon (Palmer et al. 1992), and between summer monsoon rainfall and temperatures in the eastern Indian Ocean north of Australia (Nicholls 1995), suggest that the monsoons are affected by the throughflow. Rainfall anomalies over Australia have been linked to surface temperature anomalies in the Indonesian waters, Indian Ocean, and central Pacific (Nicholls 1989). Thus, control of sea surface temperature by the throughflow implies that rainfall over Australia is affected.
A test of this hypothesis requires experiments with a coupled ocean–atmosphere model, preferably without flux correction, as the latter might compromise results. Schneider and Barnett (1997) investigated the throughflow in a simulation with such a coupled ocean–atmosphere model and found it realistic in both its transport of mass and heat. This simulation serves as a reference state for an experiment that determines changes of the climate system induced by a closure of Indonesian waters. At the outset it has to be pointed out that the thermohaline circulation in the coupled model is constrained by observed surface temperature and salinity in high latitudes and can therefore not change. However, the model is well suited to study the sensitivity to the throughflow of atmosphere and upper ocean equatorward of 60° lat, a task that also limits simulation time to a realizable effort.
The outline of the paper is a follows: section 2 describes the coupled model, and sections 3 and 4 introduce the experiment and discuss the significance of results in light of internal variability of the coupled system. Changes of the mean climate and seasonal cycle are presented in section 5 and feedbacks in section 6, followed by discussion and conclusions.
2. Coupled model
The model (ECHO, Latif et al. 1994) was developed at the Max-Planck-Institut für Meteorologie, Hamburg, Germany, and consists of an atmospheric general circulation model (ECHAM3; Roeckner et al. 1992; DKRZ 1992) and a primitive equation, ocean general circulation model. ECHAM3 has 19 levels in the vertical and is run at T42 (2.8° × 2.8°) resolution. The horizontal resolution of the ocean model is variable with latitudinal spacing of 0.5° within 10° of the equator and expanding to 2.8° resolution poleward of 20° latitude. The longitudinal resolution is 2.8° and there are 20 levels in the vertical (see Latif et al. 1994 for additional details). ECHO is a global model and couples atmosphere and ocean globally and between 60°N and 60°S without flux correction. At higher latitudes surface temperatures and salinity are additionally relaxed to climatology (cf. Levitus 1982).
A simulation of ECHO spanning 125 yr was investigated in a number of studies (Latif et al. 1994; Latif and Barnett 1994; Schneider et al. 1996; Xu et al. 1998). While the model suffers from a tendency, typical for coupled models, to split the intertropical convergence zone in the Pacific and to produce waters off South America that are too warm, it generates all major features of the atmosphere and ocean with approximately the right strength and spatial structure. The Indonesian Throughflow of the last 105 yr of this simulations with ECHO has been studied in detail by Schneider and Barnett (1997). It transports on average 13.8 Sv (Sv ≡ 106m3 s−1) and exports 0.9 PW of heat from the Pacific to the Indian Ocean, and compares favorably with observation and other simulations. Overall, the simulation of the world’s climate and throughflow by ECHO are amazingly realistic and give confidence that results reported here have some bearing for the earth’s climate.
In ECHO, passages between New Guinea and the model’s Asia and between New Guinea and Australia connect the western Pacific and the Indian Ocean. In March of year 90 of the reference run, at a time when the throughflow was near its average, these straits were closed with walls and the throughflow was forced to vanish. The model with closed throughflow was integrated for 10 yr, and the average of the last 5 yr is investigated.
In the Tropics and high latitudes this integration time allowed the coupled system to reach approximately a new equilibrium. The leading empirical orthogonal function of the 5-yr running mean of the changes of sea surface temperature (SST) after closure of the throughflow has reached steady value (Fig. 1a), explains 75% of the change, and captures nearly the entire change of SST (Fig. 1b). Higher modes, which have not reached equilibrium (Fig. 1a), have largest action in midlatitudes (Fig. 1c) but alter SST in the Tropics and high latitudes only in a minor way. At the end of the experiment, 5-yr averages of oceanic heat content of the upper 250 m are still evolving such that the cooling in the Indian Ocean and the relaxation of the zonal gradient of the thermocline in the Pacific are enhanced. In the midlatitude Pacific and Northern Atlantic, the trace of baroclinic waves is seen whose changes of the heat content are not significant, however. Thus, the duration is of sufficient length to determine the response of upper ocean and atmosphere to changes of the throughflow. Changes expected from a longer integration include eastward migrations of the Pacific warm changes, further relaxation of the zonal gradient of the thermocline in the Pacific, and an enhancement of changes in the midlatitude gyres.
A few weeks after closure of Indonesian waters, differences of simulated SST and of surface fluxes between reference integration (run REF hereafter) and integration of ECHO with closed Indonesian seas (referred to as NOTF) occur throughout the world. This is because closure of Indonesian waters not only changes the physical characteristics but perturbs the coupled ocean–atmosphere system, which then departs from the reference run due to its nonlinear nature and sensitivity to initial conditions. To distinguish effects of closure of the throughflow from internal variability of the coupled system, we compare the last 5 yr of NOTF with statistics of 105 yr of REF integration.
Significance of changes of a scalar quantity, for example SST, is determined by the rank of the average of the last 5 yr of the NOTF run in the distribution of 5-yr averages from REF. The latter are determined with 4 yr overlap, such that the 105-yr time series yield a sample size of 101. Alternatively, the Mann–Whitney test (Conover 1971) is employed, a nonparametric test that determines if the expected value of two random samples differ significantly. The test is based on the sum of ranks of the first random sample in the concatenation of both samples, and it assumes that individual samples are independent from each other. Strictly speaking, this is not true for results of the simulations since interannual and decadal signals are present. However, both tests yield very similar conclusions, and the first method, which takes into account autocorrelations, will be used in the following. The test is applied at every position, and results are mapped only if significance exceeds 95% confidence limits.
Vectors, such as the wind stress, are rotated along axis of maximum covariance of their components before significance tests are applied to each component. Significance of changes of the vector results from the product of the probabilities of each component.
Changes of annual cycles are determined by comparison of the annual amplitude and phases based on 5-yr segments of the NOTF run, and 101 5-yr segments of the REF run. Annual amplitudes and phases of REF are interpreted as polar coordinates of a vector, and significance of changes is determined using the procedure for vectors outlined above.
Comparison of NOTF run with the entire 105 yr of REF leads to rather stringent significance requirements, since changes due to throughflow closure starting from year 90 of REF are required to be significantly different from internal variability of the model, including changes on decadal timescales (Latif and Barnett 1994) and the time-varying datum they supply. This ensures that the comparison is conservative.
5. Climate sensitivity
a. Mean state
Closure of the Indonesian Throughflow causes an eastward shift of the western Pacific and eastern Indian Ocean warm pool, with accompanying changes of centers of deep convection and their expression in surface fluxes of heat and freshwater. The eastward shift of the centers of deep convection move tropical atmospheric pressure patterns associated with the Walker circulation eastward and so affect the entire Tropics. Teleconnections to midlatitudes communicate these tropical signals over the North Pacific and over the southern oceans. As a result, surface wind stress changes globally. The Pacific trades relax, while trades in the Atlantic and Indian increase.
In the southern Indian Ocean and along the eastern coast of Australia, oceanic heat content and currents reflect the changes of the Sverdrup circulation due to eastern boundary inflow into the Indian Ocean and circulation around Australia. The sensitivity of the surface wind stress to the Indonesian Throughflow causes a relaxation of spatial gradients of the heat content in the Pacific Ocean and a steepening in the Indian Ocean and, to a lesser extent, in the Atlantic. Thus, surface currents and the equatorial undercurrent weaken in the Pacific, while in the tropical Indian Ocean these are enhanced slightly. These results indicate that the control of the Indonesian throughflow on the surface winds is partially responsible for weak upwelling and vertical shear of currents in the eastern equatorial Indian Ocean.
2) Sea level
In response to closure of the throughflow, sea level increases in the Pacific, decreases in the Indian Ocean, and yields a sea level difference in excess of 30 cm between the western Pacific and the Indonesian Sea (Fig. 2). Concurrently, sea level gradients south of Australia are reduced. In the Indian Ocean sea level is decreased by more than 25 cm centered at 20°S. The wedge shape of the sea level difference is consistent with changes in surface velocity of HG and results from lack of circulation around Australia. South of Madagascar sea level decreases in the Aghulhas retroflection, again consistent with HG.
3) Surface temperature
The surface equatorial Pacific warms by up to 1 K with maximum heating east of the dateline, while Indonesian seas and eastern Indian Ocean cool by up to 0.6 K (Fig. 3). These changes represent an eastward shift of the warm pool, a relaxation of the SST gradient along the equator in the west, and an increase in the eastern Pacific. Qualitatively, this response is expected from a shutdown of the throughflow, since it transports heat from the Pacific into the Indian Ocean. In the southern Indian and Atlantic Oceans, cooler surface waters stretch from Australia to South America, while in the south Pacific, surface temperatures are increased slightly. Even though barely significant, the North Pacific off North America is warmed by up to 0.6 K, while the central North Pacific is cooled.
Over land, changes are largest over western Australia, with warming of up to 2 K. Over northeastern Asia surface temperature cooled, and there is some indication for warming over the Himalayas and Brazil.
The cooling of the eastern Indian Ocean and warming of the western Pacific are associated with an eastward shift of the centers of deep convection, as shown by changes in precipitation (Fig. 4). The eastern Indian Ocean and western Australia experience precipitation deficits of up to 3 mm day−1, while precipitation in the western Pacific increases by more than this amount. In the equatorial Pacific, precipitation increases, whereas the intertropical convergence zones are weakened. Convective precipitation dominates the signal in the Tropics and is augmented by changes in large-scale precipitation (due to large-scale convergence) of the same sign, but of half the magnitude. Over western Australia, in the Tasman Sea, and in the Pacific south of Japan large-scale precipitation changes significantly by 0.3 mm day−1 (not shown).
The convective rainfall into the western and central equatorial Pacific is accompanied by changes of short- and longwave radiation expected from the change in cloudiness (not shown). Patterns of these fields are very similar over the ocean. In the western Pacific shortwave radiation is reduced by up to 30 W m−2, whereas eastern Indian Ocean shortwave radiation is increased by almost the same amount. Longwave radiation over the ocean partially offsets shortwave changes, with increased heat gains of 6 W m−2 in the western Pacific and decreases of 4 W m−2 in the eastern Indian Ocean. Over land, in particular Australia and Siberia, increases in surface temperature are accompanied by increased longwave heat losses and increased shortwave heat gains.
6) Turbulent heat flux
Changes in the turbulent heat fluxes are dominated by the latent heat flux over the ocean and by sensible heat flux over land. Latent heat flux losses are decreased over the western Pacific and in the Indian Ocean off western Australia by less than 20 W m−2. Latent heat losses are slightly increased over the eastern Indian Ocean at 10°S and over the western Tasman Sea. Over western Australia, latent heat losses are reduced in an area where rainfall is decreased and surface temperatures are increased, and they indicate a drying of this area. Over western Australia, changes of the sensible heat flux cool by an additional 15 W m−2, and, together with increased longwave heat losses, balance heat gains due to shortwave radiation and reduce latent heat fluxes.
7) Net heat flux
The response of the net heat flux (Fig. 5) over the ocean is correlated with and counteracts changes in surface temperature; over land changes are small. This response is reminiscent of the damping influence of surface heat flux seen in the studies of El Niño (Barnett et al. 1991; Schneider and Barnett 1995) and corresponds to a heating of 10–20 W m−2 in the eastern Indian Ocean and a cooling of 10 W m−2 in the equatorial Pacific. Interesting enough, a large warming in the Tasman Sea (Fig. 3) is associated with vigorous cooling of more than 60 W m−2 and indicates large heat transfers from the ocean. Changes in the net flux are much smaller than those obtained by a Newtonian formulation in HG and indicate that feedbacks within the atmosphere moderate the surface flux response.
Changes of the net heat flux (Fig. 5) and precipitation (Fig. 4) have similar spatial patterns in the tropical eastern Indian Ocean and western Pacific. Precipitation cancels approximately two-thirds of the surface buoyancy flux due to change in net heat flux and indicates that changes of surface heat and freshwater fluxes associated with the transfer of centers of deep convection from the Indian Ocean to the Pacific alter the surface buoyancy flux little. This is in accordance to changes of the surface buoyancy flux associated with interannual migrations of the centers of deep convection (Schneider and Barnett 1995).
8) Atmospheric pressure
The eastward shift of the centers of deep convection and the warming in the central Pacific have a dramatic effect on atmospheric pressure around the globe (Fig. 6). Consistent with the approximately 90° long between the cold and warm changes on the equator (Fig. 3), atmospheric pressure response has a global wavenumber 2, with upward displacements of the 850-mb isobaric surface west of the dateline and over Brazil and much of the tropical Atlantic, and negative perturbations over the eastern Pacific, Africa, and the western Indian Ocean. The eastward shift of deep convection excites teleconnections with the midlatitudes, reminiscent of the response to interannual anomalies in the tropical Pacific (Horel and Wallace 1981; Karoly 1989). The North Pacific low is intensified, as is the Siberian high, and there is an indication of changes over North America. In the Southern Hemisphere, lows over the circumpolar current are intensified.
Further aloft, 500- and 200-mb surface show similar changes, with increasing amplitudes over the western Pacific and the Atlantic, and decreasing amplitudes over the eastern Pacific and Indian Ocean (Fig. 7). The fact that displacements of isobaric surface over the warm pool are of the same sign at 850, 500, and 200 mb suggests that this is a response to atmospheric heating due to convection.
9) Wind stress
The global changes in the atmospheric pressure due to the closure of the throughflow elicit global changes of the surface wind stress (Fig. 8). In the tropical western and central Pacific, winds converge onto the equator and are more eastward, as expected from the appearance of deep convection there, and offset the divergence of winds over the equatorial cold tongue found in the reference run (Schneider et al. 1996). Over Indonesia and in the eastern Indian Ocean, northward winds strengthen as seen in the southeast trades and the southwest winds in the Bay of Bengal. The deepening of low pressures over Brazil are accompanied by stronger southward winds from the Atlantic.
In midlatitudes, changes of surface pressure increase continental outflows over the North Pacific and northward stresses over the Gulf of Alaska. The circulation over the Atlantic is altered by a cyclonic circulation with southward stress over the eastern United States, and northward stress over the central North Atlantic. In the Southern Hemisphere, the enhancement of the lows at 50°S leads to a northward shift of the westerlies, with westward changes at 40°S and eastward changes at 60°S. Equatorward winds at the western coast of Australia are enhanced, as the eastward winds in the Tasman Sea. These changes of the wind stress suggest a possible feedback of the throughflow on its wind forcing that will be further investigated in section 6.
10) Heat content and currents
Consistent with changes of the surface temperature, closure of throughflow cools the upper-water column of the Indian Ocean and warms the Pacific (Fig. 9). The increase of heat in the Pacific is largest in the eastern half, along the equator, and under the intertropical convergence zones and indicates that the zonal slope of the thermocline in the equatorial Pacific and the ridges at the poleward edges of the countercurrents of both hemispheres are reduced. Consistent with these changes of heat content, transports of the south equatorial currents and the countercurrents are reduced (Fig. 10).
In the Atlantic, warming to the north of the equator indicates that the ridge associated with the North Equatorial Countercurrent is reduced. In addition, the South Atlantic shows a band of cooling that extends beyond the southern tip of Africa into the Indian Ocean (Fig. 9).
The heat content and circulation around Australia due to the Indonesian Throughflow and its continuation in the Indian Ocean is shown by the changes of the heat content and currents on the eastern side of Australia and across the Indian Ocean to the coast of Madagascar. North of 10°S changes of the heat content and transports suggest a strengthening of the equatorial currents and a steepening of the east–west slope of the equatorial thermocline (Fig. 9).
In addition to the changes of the transport in the upper ocean, vertical shear along the equator is sensitive to closure of the Indonesian throughflow (Fig. 11). The relaxation of zonal gradient of the thermocline along the equator in the Pacific is accompanied by a weakening of the Equatorial Undercurrent. While changes are small compared to observed values of the speed of the Equatorial Undercurrent, compared to the weak flow in the coupled model, they are of the order of 10%. In the Indian Ocean, and to a lesser extent in the Atlantic, the increase of east–west slope of the equatorial thermocline results in an increase of the vertical shear of velocity. In fact, in the eastern Indian Ocean a subsurface maximum of eastward velocity is nonexistent in the integration with open throughflow, in accordance with weak average winds and observations (Knox 1976; Knox and Anderson 1985). Closure of the throughflow results in increases of upwelling and vertical shear in the eastern equatorial Indian Ocean and partially relieves the anomalous dynamical conditions of the eastern Indian Ocean compared to the eastern Pacific and Atlantic.
b. Seasonal cycle
The seasonal cycle of heat flux of ECHO’s throughflow is smaller by a factor of 4 than the seasonal cycle of the surface heat flux in the Indian Ocean (Schneider and Barnett 1997). Oceanic changes of annual cycles due to closure of the throughflow result therefore largely from an altered mean state described above. In general, amplitudes of the annual cycle are affected, while the phase remains largely unaltered since it is tied to the solar cycle.
Changes of the annual cycle of 850-mb height are marked by a sequence of areas with alternatively larger and smaller annual cycles stretching from the North Pacific over North America into the North Atlantic (Fig. 12). These changes indicate strengthening and shifting of atmospheric pressure patterns and are reminiscent of atmospheric teleconnections of interannual changes in SST in the tropical Pacific and winter time pressure in the midlatitudes—that is, of the excitation of the PNA pattern (Horel and Wallace 1981) and, similarly, connections with the southern ocean (Karoly 1989).
Over the Indian Ocean sector, annual cycles of 850-mb height are enhanced over the eastern Indian Ocean, Indonesia and Australia, and over Asia and suggest a strengthened monsoon, consistent with the warming in the central Pacific (Palmer et al. 1992). Indeed, surface wind stress over the Bay of Bengal and Indonesian waters are significantly increased during both monsoons in the NOTF run (Fig. 13).
Changes of the seasonal cycle of SST correspond to changes of the net heat flux. They are dominated by changes in shortwave radiation in the low latitudes associated with the seasonal cycle of deep convection, and by change in the latent heat flux in midlatitude oceans.
a. Ocean–atmosphere coupling
Closure of the throughflow alters the oceanic topography and thereby affects the oceanic Sverdrup circulation. This sensitivity to the Indonesian Throughflow has been documented in experiments with an ocean model by HG. In the coupled model resulting changes of the surface temperature will alter the state of the atmosphere and of the surface fluxes and feed back on the ocean. Thus a comparison of the results of HG with the sensitivity to the closure of the throughflow in the coupled model identifies feedbacks of the coupled system.
Closure of the Indonesian Throughflow in HG alters the oceanic Sverdrup circulation in the Indian Ocean, southern Atlantic, and on the eastern coast of Australia. The response of the coupled model in these areas is very similar to results of HG; even the wedge shape of anomalous transports in the Indian Ocean that are due to the arresting of westward propagation of Rossby waves by the mean circulation (HG) are reproduced. This indicates that this response of the coupled model to closure of the throughflow results mainly from changes of the Sverdrup transport due to changes of the topography.
In the equatorial Indian Ocean, HG report no significant changes and in the Pacific away from Australia changes of steric height and currents are restricted to the equator and reflect changes of the temperature of the upwelled water. In the coupled model the sensitivity to the throughflow of SST, heat content, and currents is much greater and indicates a sensitivity through feedbacks with atmospheric forcing, most notably wind stress. Specifically, along the equator, warm SST signals centered over 170°W are associated with eastward wind stress to its east (Fig. 14) and a convergence of the meridional wind stress on the equator. Upwelling in the western Pacific weakens, and the east–west thermocline slope along the equator relaxes (Fig. 14). This together with the changes of off-equatorial curl in the intertropical convergence zones leads to the changes of the transport and vertical structure of the currents on the equator and further enhances the surface temperature changes. In the Indian Ocean, the westward wind stress is increased and leads to a shallowing of the thermocline in the west and a relative deepening in the east (Fig. 14). Thus, closure of the throughflow results in the development of a weak cold tongue in the eastern Indian Ocean. This feedback of the coupled system in the equatorial Pacific and Indian Ocean to the perturbation of the throughflow is reminiscent of the hypotheses that the delineation of warm pool and cold tongue in the equatorial Pacific results from coupled interactions (Dijkstra and Neelin 1995).
b. Throughflow forcing
The response of the coupled model gives guidance on the feedback between the throughflow and the atmospheric circulation that forces it. Since forcing of the throughflow transport by the winds is in large part described by Godfrey’s Island Rule (Godfrey 1989; Schneider and Barnett 1997), changes of the island rule transport in response to the closure are a measure for the sign of this feedback. The average wind-induced island rule transport of the reference run is 18 Sv (Schneider and Barnett 1997) and is slightly, but significantly, smaller than the wind forcing of 19.5 Sv of the NOTF experiment (Fig. 15). This suggests that a decrease of the throughflow yields an increase of its wind forcing and implies that the feedback of throughflow and island rule winds is negative and stabilizing. This in turn suggests that the interplay of the throughflow transport and winds cannot lead to self-sustained oscillations. This conclusion is not applicable to the baroclinic component of the throughflow since the latter is governed by a different set of dynamics (Schneider and Barnett 1997).
c. Warm pool temperature
Closure of the throughflow in a coupled general circulation model perturbs the oceanic heat budget in the area of highest SST—the Indo-Pacific warm pool—and results in a tendency to warm the western Pacific and cool the eastern Indian Ocean. The response of the surface fluxes to this perturbation reveals feedbacks of the coupled ocean–atmosphere system of ECHO that control, and potentially limit, highest SST. Results presented are for 5-yr averages and therefore should not show the generation of highest SST by clear skies (Waliser and Graham 1993) that have lifetimes of at most a few months (Waliser 1996).
East of 150°E changes of heat flux normalized by changes in SST (Fig. 16a) indicate that shortwave radiation has the largest cooling tendency in the high SST range. It is partially balanced by heating tendencies of longwave radiation, due to changes in opacity of the atmosphere, and, more importantly, by a reduction of latent heat loss under the centers of deep convection. This dynamic of ECHO supports the analysis and hypotheses of Zhang and Grossman (1996) that evaporation under the centers of deep convection is reduced, and is consistent with suggestions that highest SST are limited by reduction of incident radiation due to clouds (Graham and Barnett 1987; Ramanathan and Collins 1991).
There is, however, a large scatter around the median values. West of 150°E this scatter is even larger, and for the highest SST median tendencies are nil or reversed: shortwave radiation is slightly enhanced and partially balanced by decreases of longwave radiation (Fig. 16b). The areas that correspond to the high SST are found in the enclosed Indonesian waters and South China Sea, rather than the open ocean east of 150°E and suggest that the feedback of solar radiation and SST is not a universal phenomenon but may be restricted to the open ocean.
7. Discussion and conclusions
The role of the Indonesian throughflow in the global climate is investigated by comparison of coupled ocean–atmosphere model simulations with open and closed Indonesian seas. While the experiments were cast and described in terms of changes due to the closure of the throughflow, the reverse interpretation holds equally, and the discussion will be presented in terms of effects of inclusion of the throughflow in a coupled model.
The Indonesian Throughflow and its associated circulation around Australia and New Guinea deepen the thermocline in the Indian Ocean to the west of Australia; as expected from Sverdrup relation, and mass compensation shallow the thermocline everywhere else. These changes of the thermocline affect the surface in regions of upwelling or convection and result in increases of SST in the Indian Ocean and decreases of SST in the tropical Pacific (HG). This oceanic response was found by an ocean-only investigation of Hirst and Godfrey (1993) and sets the stage for the response of the coupled model. The warming in the eastern Indian Ocean shifts the warm pool to the west and drags with it centers of deep convection with their signatures of increased precipitation, reduced shortwave radiation, and convergent winds. Thus, the position of the warm pool is in part controlled by the Indonesian Throughflow. The shift of the atmospheric centers of deep convection and diabatic heating cause a global readjustment of the atmospheric pressure of wavenumber 2 in accordance to the 90° long distance between cooling in the Pacific and warming in the Indian Ocean. This global readjustment is associated with change in the surface wind stress in the Tropics. Midlatitudes are affected by atmospheric teleconnections excited by the shift of deep convection and show signature of the PNA and South Pacific patterns known from interannual anomalies. The sensitivity to the throughflow of the tropical wind stress increases horizontal gradients of heat content in the Pacific and decreases these gradients in the northern Indian Ocean and to a lesser extent in the Atlantic. An open throughflow leads thereby to an acceleration of the tropical currents including the Equatorial Undercurrent in the Pacific, while these currents are reduced in magnitude in the Indian Ocean. In fact, the eastward winds, warm conditions, deep thermocline, and weak vertical shear in the eastern equatorial Indian Ocean result in part from control of the winds by the Indonesian throughflow.
Over Australia, simulated temperature and precipitation are sensitive to the Indonesian throughflow. Qualitatively, this is expected from interannual anomalies of Australian rainfall (Nicholls 1989). However, Nicholls (1989) finds that rainfall over western Australia has a positive correlation to the SST difference between Indonesian Seas and the central Indian Ocean. Opening of the throughflow results in cooling of the Pacific, warming of the Indian Ocean, and an increase of rainfall over western Australia. This contradiction is either due to differences in SST patterns or might be an artifact of the model. Closure of the throughflow also affects the amplitude of the annual cycles, most notably midlatitude quasi stationary pressure cells and, as expected from the warming of the Pacific, the strength of the Asian monsoon.
In summary, the Indonesian Throughflow directly affects the position of the warm pool and the centers of deep convection due to its influence on the Sverdrup transport and thermocline depth. By its control on the position of the centers of deep convection, the throughflow exerts control on the tropical and midlatitude atmospheric pressure, and thereby on the surface wind stress. This latter forcing feeds back on the ocean and alters currents and the thermal structure in the entire Tropics. Thus the Indonesian Throughflow affects the climate of the entire Tropics and parts of the midlatitudes.
In addition to the determination of the sensitivity of the climate to the throughflow, closure of the throughflow perturbs the oceanic heat budget in the warm pool. The adjustment of surface heat fluxes reveals the sensitivity of the coupled system in this high SST range. In the Pacific shortwave radiation is reduced in response to an increase of SST and thus can limit growth in SST. Longwave radiation and latent heat fluxes display a positive feedback with surface temperature and can therefore not limit sea surface temperatures in the warm pool of the western Pacific. However, in the very warm waters of the South China Sea and Indonesian waters no such relationship is found. This suggests that the control of the highest SST is not universal and is, within the limits of the coupled model, restricted to the open ocean away from boundaries.
The experiment with the coupled model also determines the sign of the feedback between the throughflow and its forcing of by wind, as described by the Island Rule (Godfrey 1989). Closure of the throughflow increases wind forcing and suggests that the feedback of the throughflow transport and winds is negative. Results also only apply to the barotropic transport of the throughflow, which are governed by the island rule. The role of feedbacks of the baroclinic component of the flow and the wind cannot be addressed in this experiment.
Results have to be viewed within the limitations of the model and the experiment. The sensitivity of the coupled system to the Indonesian throughflow might be exaggerated by the strong throughflow in the reference run: its transport of 13.5 Sv is larger than most observational estimates of 5–10 Sv. The integration time of 10 yr allows only partial adjustment of midlatitude oceans and of the tropical thermocline. Hirst and Godfrey (1994) show that the adjustment of surface temperature in the eastern Indian Ocean and equatorial Pacific is rapid such that integration time is sufficient for equilibrium to be established. However, because of the slow interplay of SST, thermocline depth and the wind stress, the trend of 5-yr averages indicate that the western Pacific heat content is decreasing at the end of the coupled integration, and a longer integration of the model will display in the Pacific a further relaxation of the thermocline and the concomitant weakening of the trade winds. In midlatitudes, most notable in the Agulhas and Kuroshio region, the oceanic adjustment time is of the order of 20 yr (Hirst and Godfrey 1994) even though all patterns of change in the ocean are established after 5 yr. The stability analysis of the throughflow and its wind forcing is based on the reference run and on the run with closed throughflow—that is, two points only. It is possible that smaller variations of the throughflow transport might lead to a different sign of the feedback.
The coupled model has a few idiosyncrasies that have an unknown effect on coupled processes. It has strong numerical diffusion in the Tropics, simulates a cold tongue extending too far to the west, and has too few stratus clouds. These issues have been addressed in a new version of this model (Frey et al. 1997) and suggest that rather than extending the current simulation, it might be fruitful to repeat this experiment with the newer version of the coupled model. A remaining problem with all global coupled models is their coarse resolution that does not resolve the complicated topography and pathways in the Indonesian Seas. This issue will need to be addressed by comparison of ocean simulations with different resolutions.
Closing the Indonesian Throughflow in a coupled model is a thought experiment to determine its role in the climate system. Use of results presented here in the interpretation of variations of the coupled system has to consider that the throughflow was forced to vanish by blocking Indonesian waters. The oceanic and atmospheric response to this perturbation includes a sea level difference between the western Pacific and Indian Oceans, of 30 cm and is not the same as climate variations that reduce or bring to zero the throughflow transport. Such variations are associated with small or zero sea level difference across the Indonesian seas. Results presented might suggest that the throughflow can be used to control the westward extent of the equatorial cold tongue in a coupled model. Even though the tendency of the model to produce a split intertropical convergence zone is reduced, atmospheric pressure is degraded (Fig. 7) and argues against usage of the throughflow to tune a coupled model.
The role of the Indonesian Throughflow in interannual variability cannot be addressed with this experiment, due to its short duration and since interannual anomalies are affected both by changes of the mean state and by the lack of an oceanic connection through the Indonesian Archipelago. An investigation of the latter process requires experiments where the plumbing in the Indonesian seas allows a mean throughflow but prohibits passage on interannual signals. Model studies that selectively turn off coupling in the atmosphere in the Indian Ocean (Nagai et al. 1995; Latif and Barnett 1995) suggest that interannual anomalies in the Indian Ocean do not play an active role in the development of El Niño, and thus interannual anomalies of the throughflow should be of little consequence. Experiments with a reduced gravity model (Verschell et al. 1995) showed interannual anomalies of upper-layer thickness in the Pacific are insensitive to closure of the throughflow, while anomalies in the Indian Ocean depend critically on open Indonesian passages. This argues also against an active role of the throughflow in the development of interannual anomalies in the Pacific. However, these conclusions need to be verified by further and exciting studies with fully coupled ocean–atmosphere models.
This work has benefited from discussions with Drs. T. Barnett, N. Barth, A. Gershunov, J. Sprintall, and T. Stockdale, and would not have been possible without the enthusiasm and computational expertise of J. Ritchie. The coupled model has been generously made available by the Max-Planck-Institut für Meteorologie, Hamburg. Computations were performed at the National Energy Research Super computer Center (NERSC), and support by the Environmental Science Division of U.S. Department of Energy (DOE DE-FG03-ER61198) as part of the Atmospheric Radiation Measurement Program, and by the National Science Foundation (NSF ATM-93-14495) are gratefully acknowledged.
Corresponding author address: Dr. Niklas Schneider, Climate Research Division 0224, Scripps Institution of Oceanography, University of California—San Diego, La Jolla, CA 92093-0224.