1. Introduction
The use of simple one-dimensional heat balance equations to study the climate system, and in particular the effects of heat transport and ice–albedo feedback, goes back to Budyko (1969) and Sellers (1969). Important contributions to the theory of energy balance models (EBMs) were made by, notably, Held and Suarez (1974), North (1975a,b), and Lindzen and Farrell (1977, 1980). A thorough review of this classic literature is given by North et al. (1981). These simple models provide an elegant illustration of the ice–albedo feedback. Typically, the surface temperature is set by the competing effects of transport and radiation, while the ice cover (which sets the albedo and thus exerts a powerful control on the radiation budget) is determined by a simple threshold condition on the surface temperature. One of the hallmarks of this nonlinear interaction is the existence of at least two vastly different climates for a given solar forcing: a moderate climate with a small ice cap or no ice and a very cold, completely ice-covered climate (which we will refer to as the “snowball” state). Graphs of the ice edge versus solar constant appear in all the above-cited classic EBM papers to illustrate the multiple-valued structure of the solutions. Between the two stable equilibria lies a third solution with a large but finite ice cover. However, this large ice cap is unstable to small perturbations and is thus not a physically realizable state.
In contrast, in this paper we present a new model that supports multiple stable ice edges: a moderate climate with high-latitude ice and a colder climate with a midlatitude ice edge. We develop the new model in a series of extensions of the basic EBM, illustrating at each step how the relationship between ice edge and solar forcing changes. Sketches of the models can be found in Figs. 1 –4. As we will show, two new pieces of physics are critical to the existence of a stable midlatitude ice edge. First, the ice (which here we suppose is sea ice) must be allowed to insulate the ocean from the atmosphere. Second, the ocean heat transport must exhibit some latitudinal structure with minima at which the ice edge is found to rest. We will discuss how such structure in ocean meridional energy transport might be expected to naturally arise as a consequence of, for example, wind-driven ocean gyres.
The atmosphere and ocean play different roles in setting the sea ice edge. Studies of the modern climate indicate that the spatial distribution of ocean heat transport convergence is the most important factor setting the observed sea ice margin (Bitz et al. 2005). Recent work on modeling the Neoproterozoic snowball earth has also emphasized the critical role of ocean heat transport in limiting sea ice expansion in radically different climates (e.g., Poulsen et al. 2001; Bendtsen 2002; Lewis et al. 2003; Poulsen and Jacob 2004; Donnadieu et al. 2004). Recognition that the stability of the climate system in the face of the ice–albedo feedback might be better modeled by treating the atmosphere and ocean separately in simple models goes back at least to Held and Suarez (1974). Although there have been numerous simple model studies of interactions between sea ice and ocean thermohaline circulation (e.g., Yang and Neelin 1993, 1997; Nakamura 1996; Jayne and Marotzke 1999; Bendtsen 2002), there has been relatively little attempt to represent the wind-driven component of the ocean heat transport in such models, despite evidence from fully coupled GCMs that the wind-driven circulation is the most important barrier to sea ice expansion into the low latitudes in the approach to the snowball earth (Poulsen and Jacob 2004). Idealized models of the wind-driven ocean heat transport have been explored by Klinger (1996), Wang et al. (1995), Klinger and Marotzke (2000), Cessi (2000), Gallego and Cessi (2000), Primeau and Cessi (2001), and Hazeleger et al. (2004), among others. Our goal here is to explore the interaction of the wind-driven ocean heat transport with sea ice–albedo feedback in the framework of an EBM.
Our paper is organized as follows: in section 2 we briefly review aspects of traditional EBMs (Fig. 1), and we take the first step in our extension of them by introducing distinct representations of the atmosphere and ocean (Fig. 2). In section 3 we consider refinements that attempt to more completely capture the role of meridional structure in ocean circulation and energy transport in setting the ice edge. The resulting model is illustrated in Fig. 3. We show how the inclusion of additional physics, built into the EBM by allowing the ocean diffusivity to vary with latitude, leads to the new stable equilibrium with a large, finite ice cover. In section 4 we present a significant extension of the EBM hierarchy in which we attempt to capture not only meridional energy transport by the atmosphere but also its angular momentum transport, allowing one to infer patterns of surface winds and thence wind-driven ocean circulation and its associated energy transport. The resulting energy-momentum balance model, which is shown in Fig. 4, couples the atmosphere and ocean both dynamically and thermodynamically and also supports a large finite ice sheet as a stable solution. We present conclusions in section 5.
2. Diffusive energy balance models
a. Brief review of some classic EBM results
We have followed Sellers (1969) in representing meridional heat transport
The S0 versus ice edge curves in Fig. 1, which for brevity will henceforth be referred to as ϕi–S0 curves, illustrate the well-known multiple-equilibrium property of the EBM: for a given S0, there may be anywhere from one to five different values of ϕi that solve (1) exactly. The regions of the graph where dS0/dϕi < 0 are unstable (Cahalan and North 1979) and thus physically are unrealizable. The “large ice cap instability” occupies the subtropical and tropical latitudes; an ice margin within these latitudes expands rapidly toward the equator. The “small ice cap instability” (SICI) occupies the polar latitudes. A very small amount of polar ice must either melt completely or grow to a minimum stable size. There is thus a range of solar constants over which a total of three stable solutions coexist: ice-free, moderate ice cap, and snowball. Discussions of SICI and its physical mechanism are given by Held and Suarez (1974), Lindzen and Farrell (1977), and North (1984). The minimum stable ice cap size is proportional to δ (as can be seen in Fig. 1) and represents the minimum distance over which the temperature (and thus the outgoing radiation) can adjust locally to the albedo forcing of the ice cap—shorter length scales are wiped out by diffusion. SICI has been explored in both EBMs and more comprehensive climate models by Huang and Bowman (1992), Matteucci (1993), Lee and North (1995), Maqueda et al. (1998), and Langen and Alexeev (2004), among others. SICI is far from universal in climate models, however, as many of the above references point out. SICI can be easily eliminated from the diffusive EBM by making some small changes to the heat transport parameterization (North 1984).
Analytic methods are not practical for the more complex models considered later, so we will need to integrate the time-dependent equations numerically to map out their ϕi − S0 relationships. The basic procedure, which will be repeated for each model, is very similar to that described by Huang and Bowman (1992). We start from a warm, ice-free state and first decrease and then increase S0 by small increments, allowing the system to re-equilibrate after each perturbation, turning around just before the large ice cap instability threshold. This method finds only the stable equilibria. Numerical results for the simple EBM are plotted in Fig. 1 along with the analytic solutions; the curves differ only in the polar latitudes, where the analytic solution shows an unstable equilibrium but the numerical solution “jumps” over it (this is the effect of SICI).
In anticipation of our later results, note the uniqueness of the ice edge in Fig. 1. Although the EBM supports multiple equilibria in the form of snowball and ice-free solutions, there is never more than one stable solution with a finite ice edge.
b. A note on parameter values
Numerical values for all our calculations are listed in Table 1. Our main results are qualitative and not sensitive to specific parameter choices. For each of the four classes of models presented here, we choose parameter values giving pole-to-equator temperature differences, heat transport profiles, and ice edges in reasonable agreement with the observed climate. Two solutions are shown in each case, taking the outgoing longwave sensitivity Bout as either 2.9 or 1.7 W m−2 °C−1 and tuning Aout to give the same ϕi for S0 = 1367 W m−2. The Bout values are based on linear regression of long-term mean Northern Hemisphere National Centers for Environmental Prediction (NCEP) reanalysis OLR against temperature at 500 and 1000 hPa, respectively. Given that Ta in our models effectively represents both the surface and the midtroposphere (see below), we take these values to be reasonable upper and lower bounds on Bout. In every case, smaller Bout is associated with enhanced sensitivity of ϕi to changes in S0, analogous to increasing δ in the simple EBM (although the corresponding nondimensional parameter is not so clearly defined in the more complex models considered later).
c. Extension to a simple atmosphere–ocean EBM
Our goal here is to extend (1) to include an explicit atmosphere and ocean, with both thermodynamic and dynamic coupling. The first step is to separate the thermodynamics into two layers, with the lower layer representing the ocean and the upper layer representing the atmosphere. This is the next member of our model hierarchy, which is sketched in Fig. 2. We let the surface (ocean) and atmosphere have temperatures Ts and Ta respectively. The albedo is controlled by Ts [replacing Ta in (2b)]; the outgoing radiation is again given by Aout + BoutTa. Here Ta will be treated as a measure of the free tropospheric temperature, say at 500 hPa. However, we will express it as an equivalent surface air temperature, assuming a constant lapse rate.
Held and Suarez (1974) looked at a similar atmosphere–ocean generalization of the Budyko-type EBM. With constant coefficients, they showed a mathematical equivalence of their AO-EBM to the simple one-layer EBM. Thus, although the extension to two layers may provide a physically more meaningful framework for estimating parameter values, it does not lead to any qualitatively new behavior in the climate. Similarly, although we use the more physically motivated diffusive (Sellers-type) transport, we find that our AO-EBM [(5)] does not yield any qualitatively new behavior when the diffusivities Ka, Ko are constant in latitude.3 This equivalence of (5) and (1) is confirmed by the results plotted in Fig. 2. Despite the added complexity of the distinct atmosphere and ocean layers with separate heat transport, there is no qualitative change in the ϕi–S0 graph compared to Fig. 1 (these are numerical solutions, so only stable solutions are shown).
A clue as to what is required to endow the EBM with qualitatively new behavior is offered by Lindzen and Farrell (1977): they argue that the stability of the ice edge depends only on the range of latitudes over which a heat transport mechanism is acting to smooth temperature and the efficiency with which it does so. In the real world, the ocean is driven by the wind stress, which varies over subplanetary scales, leading to considerable structure in the ocean’s meridional energy transport. In simple EBMs we represent the ocean heat transport as a hemispheric-scale diffusion, which cannot capture the important physics of the gyre scale. In the next section we constrain the ocean in a physically more meaningful way and in doing so, we introduce smaller, more realistic scales to ocean heat transport. This turns out to be crucial to the sensitivity of the ice edge.
3. Ocean heat transport and sea ice
In this section we refine the surface Eq. (5b) of the AO-EBM to account more fully for the separate roles of oceanic transport and ice cover on the energy balance. These issues are most easily understood in an aquaplanet framework with interacting atmosphere, ocean, and sea ice but without land surfaces, as in the coupled model of Marshall et al. (2007), because land ice is constrained less directly than sea ice by sea surface conditions and ocean heat transport. We therefore now focus attention on the representation of sea ice and its interaction with the ocean.
a. Sea ice
Sea ice has three basic properties that are crucial for the large-scale energy balance: it forms where the sea surface temperature reaches the freezing point, it is highly reflective compared to open water, and it insulates the ocean surface from the atmosphere (e.g., Bendtsen 2002). The albedo parameterization in the simple AO-EBM in Fig. 2 captures the first two of these properties, but not the third. With Ko constant everywhere, there is a nonzero ocean heat transport convergence under the ice, which is communicated upward to heat the atmosphere exactly as it would be in the absence of ice.
The simplest time-dependent model of sea ice is the “0-layer” thermodynamic model of Semtner (1976), which assumes an ice top temperature Ti in balance with the heat fluxes at the surface, the temperature at the base of the ice fixed at freezing Tf , and a linear conductive heat flux through the ice pack: Fc = κi(Tf − Ti)/hi, where κi is constant. The prognostic variable is the ice thickness hi, evolving in response to imbalances in the energy flux at the top and bottom of the ice pack. This model is an approximation to the detailed thermodynamic model of Maykut and Untersteiner (1971), itself an approximation to the fundamental equations for a two-phase brine–ice mixture or “mushy layer” (Feltham et al. 2006). Bendtsen (2002) coupled this type of sea ice model to a zonally averaged atmosphere–ocean EBM.
Focusing on the steady-state response to mean annual forcing, we consider an even simpler limit, in which the ice is a perfect insulator. Setting κi = 0 in the Semtner model effectively means that the ice thickness hi drops out of the problem, and the existence of ice can be diagnosed directly from the temperatures (as in the EBM). Therefore, if we let Ti be determined by local radiative equilibrium with the atmosphere and assume that the underlying ocean temperature To = Tf everywhere under the ice, then the ocean heat transport goes to zero at the ice edge. We represent this limit with a single equation [(5b)] for a single temperature Ts simply by setting Ko = 0 and
Before proceeding to a more sophisticated ocean model, we note that simply setting Ko = 0 poleward of the ice edge but otherwise holding it constant does not generate additional equilibria in the AO-EBM [(5)]. The ϕi − S0 graph in this case (not shown) is similar to Fig. 2 except that the SICI is absent (there are stable ice edges all the way to the pole). Thus, we conclude that the insulating property of sea ice alone is not sufficient to generate new equilibria in the AO-EBM.
b. Wind-driven gyres and ocean diffusivity
In the next section we find solutions to the AO-EBM (5) with the ocean diffusivity (11) and with Ko = 0 in the ice-covered region. This model setup is sketched at the top left of Fig. 3. Note that because the ocean heat equation depends on the product CoKo, which is independent of depth hm, the steady-state solutions are also independent of hm.
c. Multiple equilibria in the AO-EBM with specified winds
To explore the properties of our new model, let us first suppose that the wind stress is externally specified. The final member of our model hierarchy, to be discussed in the next section, is capable of generating a wind stress interactively in response to the differential heating of the atmosphere. For now, we will simply “borrow” the wind field generated by the interactive model, which is plotted in Fig. 4. It features a broad band of easterlies stretching from the equator to 33° and an even broader and more intense band of westerlies peaking at 64° and extending almost to the pole. We show in the appendix that the surface stress associated with this wind field is subject to a momentum constraint ensuring that its area-weighted global integral is zero. The main disparity between this idealized wind field and the observed time- and zonal-mean zonal wind fields is the position of the westerly maximum, which actually occurs near 50°. Taking the square of the curl of this wind stress to apply our ocean diffusivity parameterization (11) leaves us with two broad regions of enhanced diffusivity centered at 32° and 76°, which we associate with subtropical and subpolar gyres; Ko has minima at the zero curl lines located near 11°, 56°, and 88°. We integrate (5) numerically with this variable Ko but with constant atmospheric diffusivity Ka. We also set Ko = 0 poleward of ϕi at each time step to set meridional ocean energy transport to zero, thus representing the insulating effect of sea ice.
Figure 3 is analogous to Figs. 1 and 2, giving results for this AO-EBM with fixed winds and insulating sea ice. Parameters are chosen to yield a high-latitude ϕi for a realistic value of S0 (see Table 1). This solution is plotted in detail in the top right of Fig. 3. The ocean heat transport
At bottom right in Fig. 3 is an additional stable solution for the same solar forcing, in which the entire subpolar gyre is frozen over and ϕi rests at the zero curl line (and the climate is consequently much colder). This new stable equilibrium state, which has no analog in the simple EBM, is the principal new result of this paper. Its existence is intimately related to the meridional structure imprinted on the ocean heat transport by the wind.
The ϕi–S0 graphs for this wind-driven model (lower left in Fig. 3) are qualitatively different than those of the constant-diffusivity cases considered earlier: there is a marked asymmetry between the cooling and warming branches due to the new multiple equilibrium regime. In the cooling phase, the ice edge advances gradually over the subpolar gyre as S0 is reduced. In the warming phase, ϕi remains at the zero curl line over a wide range of S0 values until a threshold is reached near 1400 W m−2, which causes a complete melting of the subpolar ice cover. An additional, similar multiple equilibrium regime and threshold at higher S0 values is associated with the zero curl line near the pole.
These results can be understood as follows. At steady state,
The cooling branch of the ϕi–S0 curves differ in important ways from the constant-diffusivity cases, even disregarding the existence of the multiple equilibria. Their slopes are highly variable—the ice is much more sensitive to changes in the heat budget when the edge is located in the interior of a gyre (i.e., a latitude of efficient ocean heat transport). On the other hand, ϕi is quite insensitive to changes in the heat budget when it is located near a zero curl line (i.e., at a latitude characterized by inefficient ocean heat transport). This is a somewhat more general argument than the wind-curl dependence posited here. It implies that the regions of low ocean transport efficiency are the most likely places for the ice edge to rest.
There are no qualitative differences between the small Bout and large Bout versions of the model in Fig. 3. The multiple equilibrium regime spans a larger range of S0 values for larger Bout. We emphasize that the new equilibrium is not an expression of SICI; here we have the coexistence of two finite ice caps of different sizes, and the underlying physical mechanism is rather different than the SICI mechanism outlined in section 2. This model does not have a minimum stable ice cap size (Fig. 3 shows that stable small polar ice caps are possible). Furthermore, in the new cold solution in Fig. 3 the ice edge is slaved to the position of the midlatitude zero wind curl line because there is no ocean heat transport across this line. Thus, the ice edge is fixed at a particular latitude (about 55°) for a wide range of solar forcing, unlike the multiple equilibria generated by SICI in the simple EBM (Figs. 1 and 2).
In the next section we introduce a dynamical process into the EBM in order to calculate τ interactively. As we will see, this fully coupled model generates essentially the same multiple equilibrium behavior discussed above, but without the need to externally specify a wind stress forcing.
4. An energy-momentum balance model
We now develop an extension to the atmospheric EBM framework that allows us to simultaneously represent poleward heat transport by synoptic eddies and their associated angular momentum transport that acts to maintain the surface wind stress. In so doing, and by coupling this atmosphere to the wind-driven ocean developed in the previous section, we arrive at a simple system of equations in which the atmosphere and ocean are coupled together both dynamically and thermodynamically. This final member of our model hierarchy is illustrated in the top left of Fig. 4. In it we represent the transfer of angular momentum by Reynold stresses transporting westerly momentum out of the tropics in to midlatitudes, inducing tropical trade winds and midlatitude surface westerlies. To represent this angular momentum transfer,8 we hypothesize that atmospheric eddies mix potential vorticity (PV) subject to a global angular momentum constraint—that is, eddies should not generate any net momentum but only redistribute it.
a. Formulation
Our atmosphere thus consists of two coupled diffusion equations [(13)] for PV with thermal, mechanical, and eddy forcing calculated from winds and temperature, along with the prognostic equation (15) for global mean temperature. The ocean is represented by the heat diffusion equation (5b) with the wind-driven diffusivity (11), which is now coupled to the atmosphere thermodynamically through the heat exchange Aup + Bup(Ts − Ta) and mechanically through the stress τ. We thus have a system of three prognostic PDEs and one prognostic ODE that can be integrated numerically by a simple timestepping procedure, with inversion of the QGPV according to (14) between each time step. We refer to this system as the energy-momentum balance model, or EMomBM.9 The model is readily spun up to steady state on a laptop.
Our EMomBM is similar to the zonally averaged wind-driven model discussed by Gallego and Cessi (2000), which also couples together Green’s model for atmospheric eddy momentum fluxes with an energy balance calculation and a simple description of wind-driven ocean gyres. Our model differs from this earlier work in the inclusion of sea ice and spherical geometry and the lack of an explicit delay time for the wind forcing of the ocean (in our model gyres adjust instantaneously to changes in wind forcing, implying very fast Rossby wave speeds). These differences reflect the very different intended applications of the two models. We are primarily interested in the role of the wind-driven ocean circulation in setting sea ice extent and thus, through the nonlinear albedo feedback, allowing for multiple stable equilibria (as we show in the next section). Gallego and Cessi (2000), on the other hand, focus on coupled modes of variability in the midlatitude atmosphere–ocean system, and their model exhibits decadal-scale oscillatory solutions due to the finite cross-basin Rossby wave transit time. A more comprehensive model would include both sea ice and a finite delay time for the gyres; whether such a model would exhibit multiple oscillatory solutions is left as an open question.
b. Multiple equilibria in the EMomBM
Parameters for EMomBM are listed in Table 1 (see the appendix for further definitions). Values were chosen to give roughly the same ice edge and heat transports as found for the fixed-wind AO-EBM in Fig. 3. The maximum value of atmospheric diffusivity is 6 × 106 m2 s−1, which occurs in the midlatitude lower troposphere; this is consistent with a simple scaling in terms of eddy mixing lengths and wind speeds.
The EMomBM exhibits multiple equilibria that are quite similar to that found in the fixed-wind model, as illustrated by the two solutions plotted at top right and bottom right of Fig. 4. The similarity to the solutions plotted in Fig. 3 demonstrates two things: the two-level QGPV diffusion equations [(13)] can reproduce the temperature field predicted by the one-level heat diffusion equation [(5a)], and the surface wind stress generated by the EMomBM is quite robust (the winds in the upper and lower panels look nearly the same, despite substantial changes in albedo and temperature).
Because the winds do not vary much, the ϕi–S0 graph (lower left in Fig. 4) is quite similar to its fixed-wind counterpart in Fig. 3. Thus our earlier discussion on the multiple equilibrium regimes applies equally to this EMomBM. There are, however, some differences from the fixed-wind case: the ranges of S0 for which the multiple equilibria exist are smaller, and the jump in ϕi as the system warms past its threshold is more modest. These differences are related to subtle shifts in the position and magnitude of the wind stress, and thus the shape of Ko, that occur in response to changes in ϕi (and thus the differential heating of the atmosphere). Apparently the feedback among ice, wind, and ocean heat transport in the EMomBM acts to destabilize the cold solution somewhat. Paradoxically, this may actually increase the likelihood of abrupt changes in the system under variable forcing because smaller variations in the heat budget are required to span the hysteresis loop.
Here we briefly explore the climatic implications of the hysteresis loop in the EMomBM. The ϕi–S0 graphs suggest that an external forcing that raises and lowers the energy budget of the climate system has the potential to generate asymmetric warming and cooling, without driving the climate to snowball extremes. To make this idea explicit, we integrate the EMomBM with imposed time-dependent sinusoidal variations in Aout (setting the global mean longwave cooling) while holding S0 fixed.10
The time history of the forcing and of the response for three different runs is shown in Fig. 5. We impose Aout variations on the order of 10 W m−2, which is roughly equivalent to a threefold variation in CO2 concentration, based on the classic radiative transfer calculations of Manabe and Wetherald (1967), holding relative humidity fixed. For these integrations we have set the ocean heat capacity to Co = 4 × 108 J m−2 °C−1, corresponding to an ocean mixed layer depth of about 90 m. The time scale of the forcing (2000 yr) is arbitrary but long compared to the equilibration time of the system (roughly 10 yr).
We show in Fig. 5 that by making small changes in the amplitude of Aout variations (±1 W m−2), we can generate very different climate variability in the EMomBM. Each of the three runs is initialized from a warm state and cools gradually in response to the increase in longwave emission. In one case (dashed curve) the climate varies linearly with the forcing, with the ice expanding and melting back gradually through three forcing cycles. The maximum value of Aout (first reached after 1000 yr) is not large enough in this case for ϕi to reach the midlatitude zero curl line. A second case (solid curve) does get cold enough to freeze over the entire subpolar gyre, and ϕi consequently remains fixed at the zero curl line while the “greenhouse warming” increases (Aout decreases) until the system warms past the melting threshold. The resulting climate variations have a sawtooth shape illustrating a distinct asymmetry: gradual cooling and abrupt warming. The third case (dashed–dotted curve) has a slightly greater minimum Aout value, such that the abrupt warming threshold is never reached. In this case, the climate cools gradually during the first cycle, and ϕi reaches the zero curl line and never recovers. As a result, the global mean air temperature is some 6°C colder than the other two runs at the warmest point in the cycle.
5. Conclusions
The main goal of this paper was to demonstrate the existence of an additional stable equilibrium climate state in an extension of the well-known energy balance model. The new equilibrium state features a midlatitude ice edge and has no analog in the simple EBM. It coexists with the warmer, small ice cap solution and the very cold “snowball” solution found in the simple EBM. The minimal new physics required to support this new equilibrium are twofold: the ocean heat transport must have some latitudinal structure with a minimum in midlatitudes, and the sea ice must insulate the ocean from the atmosphere (which requires that the ocean heat transport go to zero at the ice edge at steady state).
We have described a new energy-momentum balance model, at the heart of which is two new features. For the ocean, we have considered wind-driven gyres in Sverdrup balance whose meridional energy transport is represented by a heat equation in which the diffusion coefficient is set proportional to the square of the curl of the wind stress. Thus, the meridional scales of the wind forcing are imprinted on the ocean, and in particular,
Our results with the fixed-wind model indicate that the mechanical coupling between atmosphere and ocean in the EMomBM is not required to support the new equilibrium state; heat diffusion with Ko varying in latitude but fixed in time is sufficient for the new behavior. However, we feel that the EMomBM is interesting in its own right. It is one of the most compact expressions of the simultaneous conservation of heat and angular momentum in the climate system. With only a modest increase in computation over the simplest EBM, we solve for a quantity (wind stress) that is fundamental to atmosphere–ocean coupling. This EMomBM may therefore help fill in a gap in our spectrum of climate models.
The specific arguments laid out in section 3 linking
Finally, we emphasize that the hysteresis loop in our new model, with its abrupt warmings and gradual coolings, is generated in a coupled atmosphere–ocean–ice system with a purely wind-driven ocean; we do not consider any ocean overturning or thermohaline effects. This suggests a need for a more careful consideration of the role of winds in the dynamics of past climate changes—in particular, the abrupt warmings known to have occurred repeatedly throughout the last ice age.
We thank Jean-Michel Campin for advice on the numerical algorithms, David Ferreira for an insightful review of an early draft of this manuscript, and two anonymous reviewers for helpful comments. Brian Rose received support from MIT’s Climate Modeling Initiative and NSF Ocean Sciences.
REFERENCES
Bendtsen, J., 2002: Climate sensitivity to changes in solar insolation in a simple coupled climate model. Climate Dyn., 18 , 595–609.
Bitz, C., , M. Holland, , E. Hunke, , and R. Moritz, 2005: Maintenance of the sea-ice edge. J. Climate, 18 , 2903–2921.
Budyko, M., 1969: The effect of solar radiation variations on the climate of the earth. Tellus, 21 , 611–619.
Cahalan, R. F., , and G. R. North, 1979: A stability theorem for energy-balance climate models. J. Atmos. Sci., 36 , 1178–1188.
Cessi, P., 2000: Thermal feedback on wind stress as a contributing cause of climate variability. J. Climate, 13 , 232–244.
Donnadieu, Y., , G. Ramstein, , F. Fluteau, , D. Roche, , and A. Ganopolski, 2004: The impact of atmospheric and oceanic heat transports on the sea-ice-albedo instability during the Neoproterozoic. Climate Dyn., 22 , 293–306.
Enderton, D., , and J. Marshall, 2009: Explorations of atmosphere–ocean–ice climates on an aquaplanet and their meridional energy transports. J. Atmos. Sci., 66 , 1593–1611.
Feltham, D., , N. Untersteiner, , J. Wettlaufer, , and M. Worster, 2006: Sea ice is a mushy layer. Geophys. Res. Lett., 33 , L14501. doi:10.1029/2006GL026290.
Gallego, B., , and P. Cessi, 2000: Exchanges of heat and momentum between the atmosphere and the ocean: A minimal model of decadal oscillations. Climate Dyn., 16 , 479–489.
Green, J., 1970: Transfer properties of the large-scale eddies and the general circulation of the atmosphere. Quart. J. Roy. Meteor. Soc., 96 , 157–185.
Hazeleger, W., , R. Seager, , M. A. Cane, , and N. H. Naik, 2004: How can tropical Pacific Ocean heat transport vary? J. Phys. Oceanogr., 34 , 320–333.
Held, I. M., , and M. J. Suarez, 1974: Simple albedo feedback models of the icecaps. Tellus, 26 , 613–629.
Huang, J., , and K. P. Bowman, 1992: The small ice cap instability in seasonal energy balance models. Climate Dyn., 7 , 205–215.
Jayne, S. R., , and J. Marotzke, 1999: A destabilizing thermohaline circulation–atmosphere–sea ice feedback. J. Climate, 12 , 642–651.
Jeffreys, H., 1926: On the dynamics of geostrophic winds. Quart. J. Roy. Meteor. Soc., 52 , 85–104.
Klinger, B. A., 1996: A kinematic model of wind-driven meridional heat transport. J. Phys. Oceanogr., 26 , 131–135.
Klinger, B. A., , and J. Marotzke, 2000: Meridional heat transport by the subtropical cell. J. Phys. Oceanogr., 30 , 696–705.
Langen, P. L., , and V. A. Alexeev, 2004: Multiple equilibria and asymmetric climates in the CCM3 coupled to an oceanic mixed layer with thermodynamic sea ice. Geophys. Res. Lett., 31 , L04201. doi:10.1029/2003GL019039.
Lee, W-H., , and G. R. North, 1995: Small ice cap instability in the presence of fluctuations. Climate Dyn., 11 , 242–246.
Lewis, J. P., , A. J. Weaver, , S. T. Johnston, , and M. Eby, 2003: Neoproterozoic “snowball Earth”: Dynamic sea ice over a quiescent ocean. Paleoceanography, 18 , 1092. doi:10.1029/2003PA000926.
Lindzen, R. S., , and B. Farrell, 1977: Some realistic modifications of simple climate models. J. Atmos. Sci., 34 , 1487–1501.
Lindzen, R. S., , and B. Farrell, 1980: The role of polar regions in global climate, and a new parameterization of global heat transport. Mon. Wea. Rev., 108 , 2064–2079.
Mak, M., 1991: Influences of the earth’s sphericity in the quasi-geostrophic theory. J. Meteor. Soc. Japan, 69 , 497–511.
Manabe, S., , and R. T. Wetherald, 1967: Thermal equilibrium of the atmosphere with a given distribution of relative humidity. J. Atmos. Sci., 24 , 241–259.
Maqueda, M. A. M., , A. Willmott, , J. Bamber, , and M. Darby, 1998: An investigation of the small ice cap instability in the Southern Hemisphere with a coupled atmosphere–sea ice–ocean–terrestrial ice model. Climate Dyn., 14 , 329–352.
Marshall, J. C., 1981: On the parameterization of geostrophic eddies in the ocean. J. Phys. Oceanogr., 11 , 257–271.
Marshall, J. C., , and F. Molteni, 1993: Toward a dynamical understanding of planetary-scale flow regimes. J. Atmos. Sci., 50 , 1792–1818.
Marshall, J. C., , D. Ferreira, , J-M. Campin, , and D. Enderton, 2007: Mean climate and variability of the atmosphere and ocean on an aquaplanet. J. Atmos. Sci., 64 , 4270–4286.
Matteucci, G., 1993: Multiple equilibria in a zonal energy balance climate model: The thin ice cap instability. J. Geophys. Res., 98D , 18515–18526.
Maykut, G. A., , and N. Untersteiner, 1971: Some results from a time-dependent thermodynamic model of sea ice. J. Geophys. Res., 76 , 1550–1575.
Nakamura, M., 1996: Effects of ice albedo and runoff feedbacks on the thermohaline circulation. J. Climate, 9 , 1783–1794.
North, G. R., 1975a: Analytical solution to a simple climate model with diffusive heat transport. J. Atmos. Sci., 32 , 1301–1307.
North, G. R., 1975b: Theory of energy-balance climate models. J. Atmos. Sci., 32 , 2033–2043.
North, G. R., 1984: The small ice cap instability in diffusive climate models. J. Atmos. Sci., 41 , 3390–3395.
North, G. R., , R. F. Cahalan, , and J. A. Coakley, 1981: Energy balance climate models. Rev. Geophys. Space Phys., 19 , 91–121.
Phillips, N. A., 1956: The general circulation of the atmosphere: A numerical experiment. Quart. J. Roy. Meteor. Soc., 82 , 123–164.
Poulsen, C., , and R. Jacob, 2004: Factors that inhibit snowball Earth simulation. Paleoceanography, 19 , PA4021. doi:10.1029/2004PA001056.
Poulsen, C., , R. T. Pierrehumbert, , and R. Jacob, 2001: Impact of ocean dynamics on the simulation of the Neoproterozoic “snowball Earth.”. Geophys. Res. Lett., 28 , 1575–1578.
Primeau, F., , and P. Cessi, 2001: Coupling between wind-driven currents and midlatitude storm tracks. J. Climate, 14 , 1243–1261.
Sellers, W. D., 1969: A global climatic model based on the energy balance of the earth–atmosphere system. J. Appl. Meteor., 8 , 392–400.
Semtner, A. J., 1976: A model for the thermodynamic growth of sea ice in numerical investigations of climate. J. Phys. Oceanogr., 6 , 379–389.
Stone, P. H., 1978: Constraints on dynamical transports of energy on a spherical planet. Dyn. Atmos. Oceans, 2 , 123–139.
Wang, X., , P. H. Stone, , and J. Marotzke, 1995: Poleward heat transport in a barotropic ocean model. J. Phys. Oceanogr., 25 , 256–265.
White, A., 1977: The surface flow in a statistical climate model—A test of a parameterization of large-scale momentum fluxes. Quart. J. Roy. Meteor. Soc., 103 , 93–119.
Wu, G-X., , and A. White, 1986: A further study of the surface zonal flow predicted by an eddy flux parametrization scheme. Quart. J. Roy. Meteor. Soc., 112 , 1041–1056.
Yang, J., , and J. D. Neelin, 1993: Sea-ice interaction with the thermohaline circulation. Geophys. Res. Lett., 20 , 217–220.
Yang, J., , and J. D. Neelin, 1997: Decadal variability in coupled sea-ice–thermohaline circulation systems. J. Climate, 10 , 3059–3076.
APPENDIX
Derivation of the Atmospheric Model
The flow is calculated at each time step by inversion of the zonal mean PV gradients (14), subject to boundary conditions uu = ul = 0 at ϕ = 0°, 90° (an equatorial wall). The system is closed by expressing PV fluxes, heating, and stress in terms of the winds.
Plugging (12) into (A3), the PV equations take the form of diffusion equations (13). These equations are coupled together through the heating term and the dynamical constraint (A7), which sets the relative magnitudes of the diffusivities.
The classic diffusive energy balance model, (top left) illustrated in a schematic (thin gray arrows represent radiative fluxes; thick gray arrow represents poleward energy transport due to atmospheric and oceanic circulation, which is parameterized as a diffusive process). (top right) Temperature and heat transport are plotted as functions of latitude for a particular numerical solution of (1). (bottom left) We plot S0 vs ice edge ϕi for two different parameter sets yielding different sensitivities (δ = 0.32 in black and δ = 0.19 in gray; see Table 1 for all parameter values). Multiple equilibria for a given S0 can be read horizontally in these plots. The dashed curves show the analytic steady-state solution of (1) (adapted from North 1975a), including the unstable low- and high-latitude branches. The solid curves show the equilibrated numerical solutions of the time-dependent system (stable states only). (bottom right) An expanded view of the stable finite ice cover solutions. The small arrows indicate the path taken by the numerical integration; the small ice cap instability is evident poleward of about 80°.
Citation: Journal of the Atmospheric Sciences 66, 9; 10.1175/2009JAS3039.1
As in Fig. 1, but for the atmosphere–ocean EBM, Eq. (5), with constant diffusivities Ka, Ko. (right) meridional profiles of Ta, Ts,
Citation: Journal of the Atmospheric Sciences 66, 9; 10.1175/2009JAS3039.1
As in Fig. 2, but for the AO-EBM (5) with latitudinal structure in the ocean diffusivity Ko. A prescribed wind stress (taken from Fig. 4) is used to drive ocean gyres through the parameterization (11). Here Ko is set to zero under the ice to account for the insulating effect of sea ice. The gyre model gives a bimodal ocean heat transport with a minimum at the midlatitude zero curl line. This leads to qualitatively different behavior of the ice edge compared to the simple EBM. Two stable ice edges are found for a range of S0 values: (top right) one moderate solution with partial ice cover over the subpolar gyre (solution 1) and (bottom right) one cold solution with a completely ice-covered subpolar gyre (solution 2). (bottom left) The S0–ϕi curves for this model. The small arrows at lower left again indicate the path taken by the ice edge in numerical integrations.
Citation: Journal of the Atmospheric Sciences 66, 9; 10.1175/2009JAS3039.1
Solutions from the energy-momentum balance model. The atmosphere is now represented by two layers of QGPV [Eq. (13)], with the temperature Ta in thermal wind balance with the vertical wind shear. Meridional fluxes of heat and angular momentum (sketched in light and dark gray respectively) are implicitly represented by a meridional eddy QGPV flux (sketched in medium gray). The ocean is driven by τ, generated interactively by the atmospheric model (the westerlies being maintained by the convergence of momentum fluxes in midlatitudes); otherwise the ocean is identical to that shown in Fig. 3. The model supports multiple stable ice edges, as illustrated in the top right and bottom right. The S0 − ϕi relationships (lower left) are similar to those in the AO-EBM with fixed wind-driven gyres (Fig. 3).
Citation: Journal of the Atmospheric Sciences 66, 9; 10.1175/2009JAS3039.1
A time-dependent case with sinusoidally varying forcing. (top) Three separate integrations of the EMomBM are initialized from a warm, ice-free state and forced by sinusoidal variations of the longwave cooling constant Aout over 2000-yr periods. The resulting (middle) ice edge and (bottom) global mean atmospheric temperature. The three runs each produce very different climate variations as a result of small differences in the amplitude of the thermal forcing. Parameters are as in Fig. 4 (with Bout = 1.7 W m−2 °C−1), but with Co set to a more realistic value of 4 × 108 J m−2 °C−1, giving an adjustment time of about 10 yr for the system (short compared to the time scale of the forcing).
Citation: Journal of the Atmospheric Sciences 66, 9; 10.1175/2009JAS3039.1
Parameter values used for the numerical solutions of the four classes of models. In each case two solutions are computed, using two sets of values for the longwave cooling Aout, Bout as shown. The dash (—) indicates unitless parameters or parameters that are not used in particular models.
In (1) and what follows, we express the meridional component of the divergence in terms of the operator Dy[h(ϕ)] = (a cosϕ)−1∂/∂ϕ[cosϕh(ϕ)] for any function h(ϕ), where ϕ is latitude and a is the planetary radius. This operator simplifies the notation while including the necessary spherical geometric factors. Gradients are expressed in terms of a dimensional variable dy = a dϕ.
This parameter is called D in North’s papers. The equivalent parameter for the Budyko-type EBM is called δ by Held and Suarez (1974).
North (1975b) showed that the diffusive EBM and the Budyko-type EBM are equivalent under a spectral truncation. It can also be shown that the steady-state, constant-K form of (5) is isomorphic with (1) up to a spectral truncation error.
We continue to take Tf = −10°C (the canonical threshold temperature in simple EBMs, based on the mean annual position of the land-based snow line) for ease of comparison with established results. A more appropriate threshold for this sea ice model might be Tf = −2°C, roughly the freezing point of seawater. The main results of this paper (in particular, the existence of multiple stable ice edges to be discussed later) are not sensitive to this choice: we have obtained the same qualitative results using Tf = −2°C.
Here Dy is used to take the curl of stress τ (more precisely, the vertical component of the curl). This has the same mathematical form as the divergence because τ varies only in latitude.
In the real oceans the gyres contribute to poleward heat transport everywhere except in the equatorial regions, where the circulation is opposite to that of the neighboring subtropical gyres, but there is no corresponding change in sign of ΔT. However, the equatorward heat transport by the equatorial gyres is small and is overwhelmed by the substantial poleward heat transport by wind-driven subtropical cells (see, e.g., Hazeleger et al. 2004; Enderton and Marshall 2009).
It is interesting to note that Enderton and Marshall (2009) find a qualitatively similar bimodal
As has been known since the work of Jeffreys (1926), angular momentum transport by synoptic eddies cannot be represented as a diffusive process because westerly momentum is transported up the mean gradient to maintain the midlatitude westerly jet. Instead we choose to mix PV, following Green (1970).
Not to be confused with an energy–moisture balance model, which is sometimes abbreviated EMBM in the literature.
Here we choose to vary Aout partly to explore a different parameter sensitivity in the model, but we do so also because Aout is a better analog to the effects of greenhouse gas concentration than S0, since S0 projects both onto the global mean energy budget and its pole-to-equator gradient.
