1 2 3 A Theory for Polar Amplification from 4 a General Circulation Perspective 5 6 7 8 9 Sukyoung Lee 10 11 12 13 The Pennsylvania State University, University Park, PA, USA 14 Seoul National University, Seoul, Republic of Korea 15 16 17 18 1 19 Abstract 20 21 Records of the past climates show a wide range of values of the equator-­‐to-­‐pole 22 temperature gradient, with an apparent universal relationship between the temperature 23 gradient and the global-­‐mean temperature: relative to a reference climate, if the global-­‐
24 mean temperature is higher (lower), the greatest warming (cooling) occurs at the polar 25 regions. This phenomenon is known as polar amplification. Understanding this equator-­‐to-­‐
26 pole temperature gradient is fundamental to climate and general circulation, yet there is no 27 established theory from a perspective of the general circulation. Here, a general-­‐circulation-­‐
28 based theory for polar amplification is presented. Recognizing the fact that most of the 29 available potential energy (APE) in the atmosphere is untapped, this theory invokes that La-­‐
30 Niña-­‐like tropical heating can help tap APE and warm the Arctic by exciting poleward and 31 upward propagating Rossby waves. 32 33 34 35 36 37 38 39 40 2 1. INTRODUCTION 41 In spite of the complexity of the climate system, there is a recurring global-­‐scale 42 43 phenomenon known as polar amplification. Polar amplification refers to the phenomenon 44 that warming or cooling trends are strongest over high latitudes, particularly over the Arctic 45 (e.g., Manabe and Wetherald 1975; Huber et al. 2000; Rigor et al. 2000; Johannessen et al. 46 2004). Geological data suggest that during the warm periods of earth history, the equator-­‐
47 to-­‐pole temperature gradients were much smaller than the present-­‐day value. Fossilized 48 remains of frost-­‐averse plants and animals around the Arctic Ocean (e.g., McKenna 1980; 49 Tarduno et al. 1998) suggest that above-­‐freezing temperatures were not uncommon during 50 the Cretaceous and early Cenozoic. Estimates of tropical temperatures during these warm 51 periods vary, ranging from below to above the present-­‐day values (Bralower et al. 1997; 52 Douglas and Savin 1978; Pearson et al. 2001). The evidence collectively suggests that the surface equator-­‐to-­‐pole temperature 53 54 gradients were weaker during warm periods (Huber et al. 2000). Similarly, CLIMAP (Climate: 55 Long range Investigation, Mapping, and Prediction) reconstruction data show that the 56 meridional surface temperature gradient during the Last Glacial Maximum (LGM) was 57 greater than that of the present-­‐day climate. Based on this paleoclimate evidence, Budyko 58 and Izrael (1991) hypothesized that there may be a universal relationship between changes 59 in the surface meridional temperature gradient and global-­‐mean temperature change. This hypothesis was tested by Hoffert and Covey (1992), and their key figure is 60 61 reproduced here as Fig. 1. The top panel of Fig. 1 shows the latitudinal distribution of the 62 deviation of surface temperatures from the present-­‐day temperature, normalized by the 3 63 global mean of the temperature deviations. Supporting the hypothesis, the latitudinal 64 profiles are remarkably similar for the three different warm periods. As a way of testing of 65 the idea further, Hoffert and Covey (1992) constructed a least square fit to the data shown in 66 Fig . 1a (shown as a solid line), and then the resulting ‘universal’ relationship was applied to 67 the global mean temperatures of the middle Cretaceous and LGM. As can be seen in the 68 lower panel, the latitudinal distributions of the temperature predicted by this relationship 69 agrees well with the data from both time periods. This apparent universality is also evident 70 in Fig. 2 which is from more recent work by Miller et al. (2010). Across four different periods 71 (LGM, Holocene Thermal Maximum, Last Interglaciation, and middle Pliocene), there is a 72 nearly constant linear relationship between the global temperature change and the 73 corresponding Arctic temperature change. Polar amplification is also evident in the current-­‐day climate. Figure 3 shows that 74 75 fluctuations in surface temperature are small in the tropics, but they increase with latitude, 76 with largest values in the polar latitudes. Climate models in general show polar amplification 77 in response to enhanced greenhouse gas (GHG) forcing (e.g., Manabe and Stouffer 1980; 78 Hansen et al. 1984; Winton 2006; Meehl et al. 2007). The models, however, deviate rather 79 substantially in their prediction of the magnitude of the Arctic warming (Masson-­‐Delmotte 80 et al. 2006; Walsh et al. 2008). This indicates that key physical processes behind the polar 81 amplification are not well simulated by the models. Why is this phenomenon of polar amplification interesting? There are several 82 83 reasons. On the practical side, there are some obvious biological and socio-­‐economical 84 consequences. For instance, the melting of sea ice will open up new, shorter passages for 4 85 ships between Asia and Europe. Meteorologically, it can have a global scale impact. Recent 86 studies on sea ice and snow cover indicate that the melting of Arctic sea ice during the 87 summer and snow over land areas during the fall can influence the hemispheric scale 88 circulation during the following winter. These changes in the circulation are associated with 89 fluctuations in winter precipitation, blocking frequency, cold-­‐air outbreaks, and snow cover 90 over a substantial fraction of the middle latitude Northern Hemisphere (NH) (e.g., Singarayer 91 et al. 2006; Honda et al. 2009; Petoukhov and Semenov 2010; Overland and Wang 2010; Liu 92 et al. 2012). 93 In addition to the reasons described above, polar amplification is a fascinating 94 phenomenon from a theoretical point of view. Regardless of the initial cause of the warming 95 or cooling of the Arctic, the two temperature profiles (one for the Cretaceous period and the 96 other for the LGM) in Fig. 1 raise a fundamental question as to how the different equator-­‐to-­‐
97 pole temperature gradients were maintained. Consider the idealized global energy balance 98 at the top of the atmosphere (TOA), as depicted in Fig. 4a. In the tropics and subtropics, 99 there is surplus of net radiation since absorbed solar radiation exceeds outgoing long wave 100 radiation (OLR). The opposite holds in high latitudes, producing a deficit of net radiation. In 101 equilibrium, this TOA energy imbalance is compensated by poleward energy transport. 102 During the cold season when solar radiation does not reach polar latitudes, the TOA deficit in 103 polar regions is determined almost entirely by the OLR. Thus, in so far as the OLR responds 104 positively to an increasing surface temperature, the warming of the Arctic implies a stronger 5 105 poleward energy flux1 (Fig. 4b). [Of course, it is possible that Arctic OLR may actually 106 decrease as the surface warms if the Arctic becomes covered by tall, convective clouds. 107 However, this does not occur at least in the present-­‐day climate.] This leads to the following 108 question. How does the poleward energy flux increase as the climate warms? 109 In this essay, this question is addressed from the perspective of the general 110 circulation. For a recent review of observational and modeling work on the Arctic warming, 111 readers are referred to Serreze and Barry (2011). As their review paper recounts, most 112 studies on the Arctic climate have focused on local processes such as ice-­‐albedo feedback 113 (Budyko 1969). While these authors also discuss heat transport by the large-­‐scale 114 atmospheric circulation, a theoretical treatment of this topic in the context of polar 115 amplification is lacking in the literature. This essay is an attempt to advance such a theory. 116 2. THEORIES OF EQUABLE CLIMATES 117 118 a. The flux-­‐gradient relationship and baroclinic adjustment 119 Until recently, proposed mechanisms on equable climates that have considered poleward 120 heat transports include a Hadley cell which occupies the entire hemisphere (Farrell 1990), an 121 enhanced poleward oceanic heat transport (e.g., Barron et al. 1993; Sloan et al. 1995), and 122 an intensification of the thermohaline circulation due to driving by tropical cyclones (Sriver 1
One may argue that a weak temperature gradient is caused by an enhanced poleward energy transport which is in turn triggered by an initial state with a strong temperature gradient. However, in a steady state (or over time scales longer than the energy transport processes), since the poleward energy transport is not solely dependent on this presumed initial state, one is still left with the conclusion that a stronger poleward energy flux must be required to maintain an equable climate. 6 123 and Huber 2007; Korty et al. 2008). There are other proposed mechanisms that do not 124 involve poleward heat transports. These include a convection-­‐cloud radiative forcing 125 feedback (Sewall and Sloan 2004; Abbot and Tziperman 2008a,b), a vegetation-­‐climate 126 feedback (Otto-­‐Bliesner and Upchurch 1997; DeConto et al. 2000), and decreased cloud 127 reflectivity due to a reduction in the number of cloud condensation nuclei (Kump and Pollard 128 2008). But again, as was discussed above, an equilibrium climate requires an increased 129 poleward energy flux. Therefore, a mechanism that can account for the increased poleward 130 energy flux is still needed. 131 132 transport is carried out mostly by atmospheric baroclinic eddies (here eddies are defined as 133 motion that deviates from a zonal mean), it is interesting to note that none of the above 134 theories involve baroclinic eddies. There is a good reason for this. To a very good 135 approximation, the baroclinic eddy heat flux is negatively proportional to the gradient of the 136 mean temperature: 137 Given the fact that the present-­‐day extratropical (i.e, poleward of 30o) poleward heat [v’T’] = -­‐D ∂[T]/∂y, 138 139 140 where v is meridional wind, T the temperature, the square bracket an appropriately defined 141 mean, and prime the deviation form the mean. According to this flux-­‐gradient relationship, if 142 the meridional temperature gradient were to decrease, a new equilibrium state would be 143 achieved through a weaker poleward eddy heat flux. Therefore, a downgradient (poleward) 7 144 baroclinic eddy heat flux cannot account for the poleward heat transport which must 145 increase in order to maintain the weak temperature gradient of a warm climate. 146 Based on a scaling argument relevant for an idealized two-­‐layer atmosphere of 147 differing densities, Held and Larichev (1996) suggested D ~ Uλξ2 where U is the velocity wind 148 shear which is proportional to the meridional temperature gradient, λ the internal Rossby 149 radius of deformation, and ξ the supercriticality which can be thought of as the meridional 150 temperature gradient of a radiative equilibrium state. Under equable climate conditions, 151 both U and ξ decrease. According to Fig. 1, these values during the Cretaceous are roughly 152 50% of the present-­‐day values. Assuming that λ does not change the diffusivity would 153 actually be smaller under equable climate conditions. (Even if λ increases by 50% so as to 154 maintain a constant interfacial slope between the two layers – for the continuous 155 atmosphere this slope is equivalent to the isentropic slope between the equator and the 156 poles – D would still decrease.) 157 There is also the view that the statistical equilibrium state of the atmosphere 158 corresponds to the marginal state for baroclinic instability. This equilibration process, known 159 as baroclinic adjustment (Stone 1978; Lindzen et al. 1980; Cehelsky and Tung 1991), is based 160 on Charney-­‐Stern-­‐Pedlosky stability criterion (Charney and Stern 1962; Pedlosky 1964). That 161 is, if baroclinic adjustment is appliacable, the atmospheric state exhibits neutrality with 162 respect to a spontaneous growth of eddies from arbitrarily small perturbations in a stratified 163 rotating fluid. This implies that an increase in the eddy heat flux cannot take place beyond 164 that of the neutral state. Therefore, in this theory, which excludes further poleward heat flux 8 165 by finite-­‐amplitude disturbances, the reduced meridional temperature gradient of the 166 equable climate cannot be realized. None of the above two equilibration processes is caused by an exhaustion of zonal-­‐
167 168 mean available potential energy (ZAPE; Lorenz 1955). In the present-­‐day atmosphere, eddy 169 available potential energy is only about 10% of the zonal-­‐mean available potential energy 170 (Peixoto and Oort 1994; Li et al. 2007). According to the above equilibration theories, the 171 baroclinic eddy heat flux is either saturated with respect to linear instability or constrained 172 in the context of homogeneous baroclinic turbulence (Salmon 1980; Vallis 1988; Held and 173 Larichev 1996). In reality, however, the atmosphere is neither devoid of finite amplitude 174 disturbances nor does it resemble homogeneous turbulence: for example, the atmosphere is 175 almost always subject to finite-­‐amplitude disturbances from the tropics (think of El Niño-­‐
176 Southern Oscillation or the Madden-­‐Julian Oscillation (MJO)), and there is inhomogeneity 177 introduced by orography and the land-­‐sea distribution. Thus, if, somehow, the mostly 178 untapped ZAPE can be released through these forcings, in principle, enhanced poleward 179 heat transport, required to maintain the equable climates, can be realized. Of course, this is not to say that the contributions from oceanic heat transport can be 180 181 ignored. The point is that shortcomings in the above equilibration theories should not 182 prevent us from searching for other atmospheric processes that can contribute to the 183 maintenance of an equable climate. Such a theory is needed because oceanic heat transport 184 alone cannot explain the high-­‐latitude continental winters of above-­‐freezing temperatures 185 of the Cretaceous and early Cenozoic. In fact, the hemisphere-­‐wide Hadley cell (Farrell 186 1990), whose sinking branch occurs in the Arctic, can warm high-­‐latitude continental interior. 9 187 However, this theory requires a tropopause height of ~ 30 km, 2-­‐3 times that of the present-­‐
188 day value. In this essay, we present another plausible theory for equable climates which 189 involves a circulation that is more similar to that of the present-­‐day climate. 190 b. Criteria for an alternative mechanism 191 192 Is there an alternative mechanism that (1) can help explain equable climates, but (2) 193 does not require a radically different atmospheric state, (3) does not involve either the flux-­‐
194 gradient relationship or baroclinic adjustment, and possibly (4) accounts for the apparent 195 universal relationship? As was mentioned earlier, given the fact that only a small fraction of 196 the ZAPE is being tapped, processes that can tap the remaining vast storage of ZAPE may 197 potentially satisfy all of the four conditions. Is there such a process? A hint can be seen from present-­‐day stationary waves. During the boreal winter, 198 199 poleward of ~40oN, the poleward transport of moist static energy is dominated by its 200 stationary wave contribution (Oort and Peixoto 1992). This can be accounted for by linear 201 wave theory: vertically propagating Rossby waves, as signified by upward and westward tilt 202 in the pressure (or geopotential) field, transport heat poleward. According to the linear 203 theory of Charney and Drazin (1961), vertical wave propagation is possible for waves with 204 large horizontal scales embedded in a weak westerly wind. The NH winter circulation is in 205 fact most favorable to this vertical propagation. Stationary waves are forced by both orography and heating, although these two 206 207 forcings are not independent of each other. For example, orographic forcing is dependent 208 on pressure difference across the orography, and the pressure field may be influenced by a 10 209 heat-­‐induced circulation. Likewise an orographically-­‐forced circulation can influence heating 210 by affecting air temperatures over the ocean, hence the ensuing convection and convective 211 heating. Since these `forced tappings’ of ZAPE are independent of baroclinity and instability, 212 this mechanism satisfies the condition (3). Stationary waves are not the only agent of a forced tapping. For instance, the 30-­‐60 213 214 day tropical convective system, known as Madden-­‐Julian Oscillation (MJO; Madden and 215 Julian 1971, 1972), can excite poleward propagating Rossby waves (Matthews et al. 2004). 216 Sardeshmukh and Hoskins (1986) showed that waves excited by tropical convective heating 217 can escape into the extratropics through a subtropical Rossby wave source. In the NH, this 218 subtropical Rossby wave source is located over southeast Asia and the adjacent western 219 subtropical Pacific Ocean where the convective divergent flow from the Indo-­‐western Pacific 220 warm pool (simply `warm pool’ hereafter) interacts with the tight absolute vorticity gradient 221 associated with the subtropical jet. The wave that emanates from the Rossby wave source 222 can propagate poleward until it encounters a turning latitude. The larger the zonal scale of 223 the wave, the farther it can propagate into high latitudes (Hoskins and Karoly 1981). 224 Therefore, when MJO convection is over the warm pool region, the Rossby wave excited by 225 this MJO heating constructively interferes with the stationary wave forced by climatological 226 warm pool convection (Garfinkel et al. 2010; 2012). This constructive interference can 227 strengthen the forced tapping, hence enhance poleward heat transport. Thus, stationary or 228 non-­‐stationary, large horizontal scale (zonal wave numbers 1 or 2) tropical convective 229 heating can tap ZAPE. This possibility is depicted in Fig. 5a. (Figure 5a also shows the 230 transient baroclinic eddy heat flux. This will be discussed in Section 3.) 11 231 The forced tapping of ZAPE can also arise from an eddy momentum flux which is 232 strongest in the upper troposphere. Since the potential vorticity gradient is positive in the 233 upper troposphere, in the latitude of the wave source, or wave stirring, there is eastward 234 acceleration by an eddy momentum flux convergence. Likewise, in the latitude of a wave 235 sink or wave dissipation, there is a westward acceleration by an eddy momentum flux 236 divergence. Thus, in the stirring (dissipation) region, the vertical shear of zonal wind 237 increases (decreases). To maintain thermal wind balance, the meridional temperature 238 gradient must increase (decrease) across the stirring (dissipation) region. Thus, poleward 239 (equatorward) of the dissipation region, there must be a sinking (rising) motion so as to 240 bring adiabatic warming (cooling). In the atmosphere, the Hadley cell is driven in part by 241 wave dissipation in the subtropics (Pfeffer 1982; Kim and Lee 2002; Walker and Schneider 242 2006); by the same token, in midlatitudes where most of the stirring occurs (due to 243 baroclinic eddies), there is the thermally indirect Ferrel cell. In the same way, if a forced wave, say, excited by tropical convective heating, 244 245 manages to propagate into the extratropics and dissipates in mid-­‐to-­‐high latitudes, there 246 should be adiabatic warming poleward of these latitudes. Figure 5b illustrates this 247 possibility: A tropical convective heat source excites a Rossby wave, and as this Rossby wave 248 propagates poleward, westerly momentum is transferred into the wave source region 249 (indicated by the symbol, ⊙, in Fig. 5b). If this momentum flux convergence is sufficiently 250 strong, the equatorial wind becomes westerly, i.e., the circulation evolves into a 251 superrotating state. If the angular momentum of the atmosphere is to be conserved, the 252 eastward acceleration in the tropics must be balanced by a westward acceleration elsewhere. 12 253 From the arguments presented above, this westward acceleration will take place in the 254 region of wave dissipation. This region is indicated by the symbol, ⊗, in Fig. 5b. As 255 mentioned above, thermal wind adjustment requires that this momentum sink produce 256 downward motion on its poleward side. The downward motion compresses the air and 257 warms it adiabatically. Therefore, if this downward motion occurs in polar regions, a 258 localized tropical heat source can warm high latitudes. In the atmosphere, such a localized 259 heat source can be found over the warm pool. If this process were to intensify as the climate 260 warms, then the Rossby wave response could be an important contributor toward polar 261 amplification. Thus, if this tropical stirring mechanism is responsible for polar amplification, 262 one can expect that the resulting equable climate must be accompanied by an equatorial 263 upper tropospheric wind which is less easterly or even superrotating. 264 265 found in a modeling study of equatorial superrotation. With an idealized two-­‐layer GCM, 266 Saravanan (1993) examined the transition from a ‘normal’ state to a superrotating state by 267 subjecting the model atmosphere to zonal-­‐wavenumber-­‐two heating on the equator. The 268 transition takes place as the heat source is gradually strengthened. Because the zonal mean 269 of the heat source is zero, the zonal mean ‘radiative equilibrium’ temperature field is 270 identical between the normal and superrotating states. In spite of their identical radiative 271 equilibrium temperature field, Fig. 1e of Saravanan (1993) shows that the midlatitude eddy 272 heat flux of the superrotating state is only about half that of the normal state. Although not 273 discussed in that study, according to the flux-­‐gradient relationship, this implies that the 274 meridional temperature gradient of the superrotating state should be weaker than that of Indeed, one of the most spectacular examples of such a forced wave effect can be 13 275 the normal state. This, in turn, implies that there must be a process, other than the eddy 276 heat flux, that warms high latitudes and/or cools the tropics in that model. Because there is 277 no ocean in the model, the high-­‐latitude warming must take place through an atmospheric 278 process, most likely through adiabatic warming. Although this result is from a highly 279 idealized model, it nevertheless alludes to the possibility that an increase in strength and a 280 localization of tropical convection may contribute toward polar amplification. 281 282 explain the universal relationship between the global mean temperature and the meridional 283 temperature gradient, there must be a process where localized warming (cooling) promotes 284 (hinders) the forced tapping. Given the theoretical expectation that an enhanced large-­‐scale, 285 localized tropical convection (acting as wave stirring) can cause winter Arctic warming, if 286 GHG warming can intensify the existing zonally localized convection (currently the strongest 287 convection occurs over the warm pool), then condition (4) can be satisfied. In other words, if 288 the GHG warming intensifies the existing east-­‐west gradient in tropical convection, according 289 to the theory presented here, poleward heat transport will intensify, hence the Arctic will 290 warm more than the tropics. Here, the zonal localization is emphasized because the zonally 291 symmetric part of the tropical heating does not excite the forced waves. 292 Now turning our attention to condition (4), if these forced tapping processes were to 293 c. Does the west-­‐east gradient in the tropical convection increase with warming? 294 In the current climate, it is uncertain whether or not the zonal asymmetry in 295 convective heating over the tropical Pacific would intensify in a warming world. In fact, this 296 has been a subject of debate in the climate science community in the context of the Walker 14 297 Circulation which is characterized by rising motion over the warm pool region and sinking 298 motion over the eastern tropical Pacific. Most climate model simulations of the twentieth 299 century show a weakened Walker circulation (Vecchi et al. 2006; Vecchi and Soden 2007), 300 and this model response has been attributed to GHG warming (Held and Soden 2006). 301 However, the opposite trend has been found in satellite data analysis (Sohn and Park 2010), 302 and also in reanalysis, reconstruction, and in-­‐situ measurements (L’Heureaux et al. 2013). 303 Since a trend toward rising (sinking) motion coincides with increased (decreased) SST and 304 active (inactive) convection, the Walker Circulation trend has often been equated with 305 trends in SST trend and/or convective precipitation. However, this is not necessarily the case. 306 The experiment by Clement et al. (1997) shows that when their model’s SST is uniformly 307 raised, while keeping the trade wind (the surface branch of the Walker circulation) speed 308 constant, the model responds by increasing the east-­‐west SST contrast, producing a La-­‐Niña-­‐
309 like SST field. Similarly, based on the thermodynamic energy balance in the tropics, one can 310 expect that the Walker circulation may weaken while the east-­‐west contrast in convective 311 heating increases; this is possible if the fractional increase in static stability is greater than 312 that of the convective heating. In paleoclimate studies, there is emerging evidence that during times when the climate 313 314 was warm (cold), the tropical SST took on a La-­‐Nina-­‐like (El-­‐Nino-­‐like) spatial structure. Stott 315 et al. (2002) analyzed seawater temperature and salinity reconstruction data for the late 316 Pleistocene, and found that during interstadials (periods of warm high latitudes) the tropical 317 Pacific warm pool was fresher, implying enhanced convection, while the stadials (periods of 318 cold high latitudes) coincide with a saltier warm pool, suggestive of suppressed convection. 15 319 This finding is at odds with the conclusions of an earlier study by Lea et al. (2000), but Stott 320 et al. (2002) pointed out that the site used by Lea et al. (2000) may be too far to the east to 321 be considered as representative of the western warm pool. The association of a warm (cold) 322 climate with La-­‐Niña-­‐like (El-­‐Niño-­‐like) conditions is reported in other studies such as 323 Koutavas et al. (2002) who showed that the LGM coincided with El-­‐Niño-­‐like conditions in 324 the tropical Pacific, and Visser et al. (2003) who found that warm pool SST increased by 3.5-­‐
325 4.0oC during the last two glacial to interglacial transitions. 326 3. TESTING OF THE FORCED TAPPING MECHANISM 327 328 The forced tapping mechanism was tested with a GCM experiment (Lee et al. 2011a) in the 329 context of the equable climate of the Mesozoic and Cenozoic2 when high-­‐latitude 330 continental winters are believed to have undergone above-­‐freezing temperatures (Huber 331 2008; Spicer et al. 2008). In a series of model runs, a localized heat source was prescribed in 332 the western corner of the tropical Pacific ocean (see the red box in Fig. 6a) where warm pool 333 convection was postulated to have existed by the authors; elsewhere in the tropics, the 334 same amount of heating is subtracted (see the blue box in Fig. 6a) so that the zonal mean 335 net heating in the tropics is zero. Figure 6a shows the difference in January surface 336 temperature between a run with 150 W/m2 contrast (between the red and blue regions) and 337 a run with no heating contrast. Although the zonal mean diabatic heating is identical 338 between the two runs, the run with the tropical heat perturbation produces Arctic 339 temperatures that are as large as 16oC warmer. Figure 6b shows evidence of poleward 2 The Mesozoic and Cenozoic range from about 145 to 50 million years ago. 16 340 propagating Rossby wave trains emanating from the heating region in the tropics, supporting 341 the hypothesis that large-­‐scale, localized tropical heating can warm the winter Arctic by 342 exciting poleward propagating Rossby waves. As the tropical heat perturbation intensifies, the Arctic surface air temperature 343 344 exhibits a marked increase (Fig. 7b), yet the temperature in the equatorial region (Fig. 7a) 345 remains close to being constant. This behavior is reminiscent of the ‘universal’ relationship 346 depicted in Figure 1. It is unknown whether there was a warm pool during the Mesozoic and 347 Cenozoic, nor how strong was the convective heating. However, this proof-­‐of-­‐concept 348 calculation shows that stirring of the tropical atmosphere (by way of a heat perturbation) 349 can produce cold season polar amplification with a muted equatorial temperature change. 350 351 found that between 40oN and 60oN, the warming is due to the stationary eddy heat flux (the 352 prescribed tropical heating is stationary); between 60oN and 80oN the warming is by the 353 transient eddy heat flux (see Fig. 5a; since the imposed heat perturbation is stationary, the 354 transient eddy heat flux should be regarded as a response to the stationary eddy heat flux), 355 and between 70oN and 80oN, it is caused by adiabatic warming driven by stationary wave 356 momentum flux. It turned out that poleward of 80oN, the warming is caused by downward 357 infrared radiation (IR) associated with increased cloud cover. Because the ultimate driver of 358 these responses is the stationary tropical heat perturbation, the logical conclusion is that 359 dynamical processes -­‐ meridional heat flux and adiabatic warming -­‐ first warms the Arctic 360 Ocean, and the transport of water vapor and cloud liquid water (by the tropically forced 361 Rossby wave, and perhaps by the transient eddies that respond to the heat flux convergence Does this polar amplification occur as was theorized in Section 2b? Lee et al. (2011b) 17 362 by the forced wave) into the Arctic leads to further warming through enhanced downward 363 IR; once it encounters the cold Arctic air, because of the low saturation vapor pressure, the 364 water vapor condenses (latent heat release) to form clouds, and as cloud cover increases, 365 radiative forcing by the clouds can take over the warming. This final step in the process is 366 sketched in Fig. 5c. 367 368 Arctic winter, and this is consistent with the La-­‐Niña-­‐like trend of intensified tropical 369 convection over the warm pool region. Lee et al. (2011b) found with global European Center 370 for Medium-­‐Range Weather Forecasts reanalysis data (ERA-­‐40) that winter Arctic warming 371 between 1958-­‐2001 is contributed by increases in the frequency of occurrence of a few 372 teleconnection patterns; these patterns grow and decay at intra-­‐seasonal time scales where 373 the growth of aggregation of these patterns is preceded by enhanced warm pool convection 374 and the decay is followed by downward IR over the Arctic Ocean. Poleward eddy heat flux 375 and adiabatic warming also contribute to the Arctic warming trend. Because it takes 7-­‐10 376 days for a Rossby wave to propagate from the tropics to high latitudes (Hoskins and Karoly 377 1981), the mechanism described here, if it is relevant, should be operating on intraseasonal 378 time scales. This is found to be the case in Lee et al. (2011b). This physical process is found to contribute to the present-­‐day warming trend in the 379 The connection between localized tropical convection and Arctic warming at 380 intraseasonal time scales is highlighted by the findings of Yoo et al. (2011). Using ERA-­‐
381 Interim data, they showed that when the MJO convection enhances (suppresses) the warm 382 pool convection, Arctic warming (cooling) follows. Here, it was again found that the positive 383 warm pool convection anomaly is followed by dynamic warming, then by moisture transport 18 384 into the Arctic, and an increase in downward IR (Yoo et al. 2012a). The implied causal 385 relationship between the convection and the Arctic warming was supported by model 386 calculations (Yoo et al. 2012b). This warm pool convection-­‐Arctic warming linkage is also 387 present at interannual time scales, since La-­‐Niña (El-­‐Niño) years are accompanied by 388 enhanced (suppressed) poleward moist energy flux in the extratropics and warmer (cooler) 389 Arctic (Lee 2012). To emphasize the linkage between the tropical convection and Arctic 390 warming, Lee (2012) named this forced tapping process as the Tropically Excited Arctic 391 warming (TEAM) mechanism3. 392 393 4. MORE ON POLEWARD HEAT FLUX AND DOWNWARD IR 394 Without invoking the TEAM mechanism, some recent studies have questioned the 395 importance of the ice-­‐albedo feedback, and have instead stated that a poleward 396 atmospheric heat (or moist static energy) transport may be the key process behind the Arctic 397 amplification (Alexeev 2005; Lu and Cai 2010; Graversen 2007; Graversen and Wang 2009). 398 Held and Soden (2006) showed that the fourth IPCC assessment report (AR4) models, under 399 Special Report on Emissions Scenarios (SRES) A1B forcing scenario, produce an increase in 400 the poleward atmospheric heat flux, while the changes in the oceanic heat transport are 401 negligible. Consistent with this increase in poleward heat flux, Wu et al. (2010) and Zelinka 3 The Pacific Decadal Oscillation (PDO) may be regarded as another phenomenon which can be used to test the TEAM mechanism since the associated SST pattern shares a resemblance with that of El Nino. However, care needs to be taken in doing so because it is the tropical convection pattern that matters in the TEAM mechanism; during the positive phase of the PDO (El-­‐Nino-­‐like SST) which started in late 1970s, the hydrographic salinity record indicates a continued freshening in the western warm pool region (Durack and Wijeffels 2010), suggesting that there has been a La-­‐Nina-­‐like trend in convective precipitation. 19 402 and Hartmann (2012) showed that the equator-­‐to-­‐pole gradient in the TOA net radiation 403 strengthens in response to GHG-­‐driven warming. Analyzing the time mean states of the 404 models, both studies found that the increase in the TOA radiation gradient is caused by 405 water vapor and cloud feedbacks. Wu et al. (2010) interpreted that these feedback processes increase the baroclinicity, 406 407 and that the transient eddies respond to this increase in baroclinicity. However, their Fig. 5 408 shows that the match between baroclinicity and the poleward eddy flux is poor: the increase 409 in baroclinicity occurs mostly in the upper troposphere, yet the increase in eddy heat flux is 410 largest in the lower troposphere. Recognizing this discrepancy, they suggested that eddies 411 are influenced by the upper tropospheric baroclinicity. Even though the mechanism that 412 drives this process remains unclear, it appears that this view, that an increase in the 413 poleward eddy heat transport is a response to a strengthening of the equator-­‐to-­‐pole 414 gradient in the TOA net radiation, or upper tropospheric baroclinity, is shared by others (Lu 415 and Cai 2010; Zelinka and Hartmann 2012). The prevalence of this view is understandable, given the importance of baroclinic 416 417 instability and the flux-­‐gradient relationship in our understanding of the atmospheric 418 circulation. However, it is not feasible to tease apart causal relationships from time mean 419 states. Instead of the causality proposed by the above studies, it is also possible that a 420 strengthening of the poleward energy flux, for instance, driven by tropical stirring (through 421 localized convection), i.e., the TEAM mechanism, causes the TOA net radiation gradient to 422 increase. If one were to diagnose the two model runs in Lee et al. (2011a), without 20 423 knowledge of how the model was perturbed, one may conclude that the clouds, by 424 influencing the TOA net radiation gradient, caused the energy transport to strengthen. The AR4 model studies also found that downward IR, caused by increased moisture 425 426 and cloud cover, plays an important role for warming the Arctic (e.g., Winton 2006). Two 427 different processes can contribute to this increase in downward IR: a local moisture 428 feedback process (Abbot et al. 2009) and remote process which invokes moisture transport 429 from outside of the Arctic. The increase in poleward latent heat flux in the AR4 models (Held 430 and Soden 2006; Wu et al. 2010) suggests that the remote process plays an important role 431 for enhancing the downward IR. 432 5. SUMMARY, CAVEATS AND FURTHER QUESTIONS 433 From the perspective of the atmospheric general circulation, this article advances a 434 435 conjecture that an equable climate, which is characterized by a weak meridional 436 temperature gradient, can be realized if the vast storage of zonal available potential energy 437 (ZAPE) can be tapped with a forcing which relies on neither baroclinic instability nor the flux-­‐
438 gradient relationship. Under usual circumstances, localized tropical convection is perhaps 439 the most viable candidate for such a forcing. From this perspective, a theory, referred to as 440 the TEAM (Tropically Excited Arctic warMing) mechanism, has been put forward (Lee 2011a, 441 b; Yoo et al. 2011, 2012a, b; Lee 2012) where the main thesis is that an enhancement in the 442 warm pool tropical convection strengthens and/or more frequently excites poleward 443 propagating Rossby waves, and that the wave dynamics and subsequent increase in 444 downward infrared radiation lead to warming over the Arctic. 21 This general-­‐circulation based theory can help explain: (1) an enhanced poleward 445 446 heat transport, conforming to the TOA energy balance; (2) that Arctic warming occurs during 447 the winter when solar radiation does not reach the Arctic; (3) the paleo-­‐equable climate 448 warmth in the high-­‐latitude continental interior; (4) the apparent universal relationship 449 between global-­‐mean temperature and the meridional temperature gradient; (5) the 450 apparent invariance of equatorial temperature in the widely varying climatic conditions (Fig. 451 1); (6) the tendency of models to produce equatorial superrotation when the model climate 452 is warmed (Lee 1999; Huang et al. 2001; Cabarello and Huber 2010). Regarding (2), the leading theory has been that the ice-­‐albedo feedback process 453 454 (Budyko 1969) warms the Arctic Ocean during the warm season and then release the heat 455 during the following winter (Serreze and Barry 2011). Figure 8a is adopted from Stroeve et 456 al. (2011), and shows a schematic description of the ice-­‐albedo feedback process in more 457 detail. In this framework, the GHG-­‐driven warmer air thins the spring sea ice and melts the 458 summer sea ice, with latter process being amplified by ice-­‐albedo feedback. The warmer 459 ocean then releases the stored heat in the following fall and winter, a process which hinders 460 sea ice formation and thickening. However, recent studies have raised questions on this 461 picture: Over regions covered by sea ice, the surface during the cold season is actually 462 warmed by a heat flux from the atmosphere (Graversen and Wang 2009; Lee et al. 2011a); 463 Park et al. (2013) and Flourney et al. (2013) found that during winter and early spring, 464 positive downward IR anomalies over the Arctic are preceded by anomalously strong warm 465 pool convection, and followed by a statistically significant reduction in Arctic sea ice; Persson 466 (2012) analyzed measurements from the Surface Heat Flux of the Arctic Ocean (SHEBA) 22 467 experiment and found that melt onset is preceded by springtime free-­‐tropospheric warming, 468 and is punctuated by atmospheric synoptic events. The autocorrelation of Arctic sea ice further suggests that winter processes such as 469 470 the TEAM mechanism may play more active role than was previously thought. Figure 9 471 shows that the April sea-­‐ice area (its lag correlations are highlighted by the first vertical line 472 in Fig. 9) is uncorrelated with the previous melting season (May-­‐September), but is 473 correlated with the previous cold season (November – March). The April sea-­‐ice area is also 474 correlated with the sea-­‐ice area for much of the following year, until the month of March. In 475 contrast, the September sea-­‐ice area (highlighted by the second vertical line in Fig. 9) is 476 significantly correlated with all of the previous months’ sea-­‐ice area, but for those months 477 with positive lags, statistically significant correlations are found only until the following 478 March. One possible explanation for these lag correlations is that winter climate conditions 479 may be more important than summer conditions. In fact, Persson (2012) reports for a 480 synoptic weather event that during the SHEBA experiment the ice-­‐albedo feedback effect 481 was much weaker than warming by downward IR. Graversen and Wang (2009) also found in 482 their model experiment that the ice-­‐albedo effect is weaker than the downward IR effect. 483 They pointed out that this is consistent with the fact that the Arctic is covered by clouds 484 most of the time (Wang and Key 2005). Therefore, contrary to the view that summer season 485 warming is the main driver, it is possible that the winter warming, perhaps driven by the 486 TEAM mechanism, may act as the main player by its hindering of the formation of sea ice, 487 and thus by promoting summer sea ice melting. This possibility is presented schematically in 488 Fig. 8b. 23 Point (4) above also needs further discussion. As indicated by Miller et al. (2010), the 489 490 apparent universal relationship (Figs. 1 and 2) may be misleading since the forcing 491 mechanisms vary among the different time periods. On the other hand, there is emerging 492 evidence that La-­‐Nina-­‐like conditions prevailed during warm periods such as Pliocene 493 (Rickaby and Halloran 2005), interglacial periods (Visser et al. 2003), and interstadials during 494 late Pleistocene (Stott et al. 2002); whereas El-­‐Nino-­‐like conditions were present during cold 495 periods such as stadials (Stott et al. 2002), glacial periods (Visser et al. 2003), and LGM 496 (Koutavas et al. 2002). Therefore, it is possible that the Arctic amplification, whether it is 497 ultimately caused by GHG or by orbital forcing (e.g., HTM and LGM were forced by variations 498 in orbital parameters), may be realized through the influence of these forcings on tropical 499 convection. There are also caveats to the TEAM mechanism. Although the premise of the theory 500 501 hinges on the forced tapping of ZAPE by dynamical warming, it turns out that the tropical 502 warm pool convection also causes downward IR to increase over the Arctic, and this process 503 appears to have an even greater impact on the Arctic surface temperature than does the 504 dynamic warming (Lee et al. 2011a,b; Yoo et al. 2012b). Hence, for the TEAM mechanism to 505 be effective, moisture transport into the Arctic (and subsequent condensation into cloud 506 liquid water droplets) is critical. Since the moisture content is higher in warm air than in cold 507 air, it is to be expected that poleward sensible heat flux caused by the forced tapping would 508 be also accompanied by an enhanced moisture flux. Nevertheless, this downward IR is not 509 part of the untapped ZAPE in the sense of Lorenz (1955), although it is in the sense of 510 available energy formulated by Bannon (2012). 24 Secondly, it is unclear whether dynamic warming can occur over the Arctic under 511 512 different climatic conditions. Adiabatic warming is an element of the TEAM mechanism, but 513 the location of the downward motion (thus the adiabatic warming) depends on where the 514 westward acceleration occurs (the location indicated by ⊗ in Fig. 5b), and also on the 515 horizontal scale of the overturning circulation which depends on Rossby radius of 516 deformation. Therefore, there is no guarantee that the downward motion would occur over 517 polar regions under a different basic state or during a different time period. Similarly, even 518 though poleward propagating waves can also transport heat poleward, since the poleward 519 heat flux can occur only if waves can propagate upward and poleward, and since this in turn 520 depends on the background state (meridional and vertical refractive indices for Rossby 521 waves are dependent on the background state), it is uncertain whether eddy heat flux 522 convergence, driven by tropical convection, would occur under a different basic state and 523 time period. Thirdly, the TOA energy balance of equable climates warrants further discussion. One 524 525 premise of the TEAM mechanism is that an equable climate arises in part from a greater 526 Rossby-­‐wave-­‐driven poleward water vapor and cloud liquid water transport which leads to 527 an increase in cloudiness and thus a larger outgoing IR in polar regions. The resulting 528 increase in the meridional gradient of the TOA net radiative flux by these waves requires 529 that there be a stronger poleward energy transport. However, this may not be the case if 530 the Arctic is covered by optically thick, tall clouds, similar to the convective clouds in the 531 tropics. In this case, the cloud tops may be sufficiently cold that the TOA radiative deficit 532 over the Arctic may be small enough for the meridional gradient in the TOA radiative flux to 25 533 be greater than that of a colder climate. It could be that in the initial stage of warming, Arctic 534 amplification occurs through an enhanced poleward energy flux, such as the TEAM 535 mechanism, and that as the warming continues and sea ice melts away, convection may 536 ensue as shown by Abbot et al. (2009). In closing, we reiterate that the TEAM theory is incomplete, but that progress can be 537 538 made in the topic of polar amplification and equable climates by going back to basics and 539 considering the general circulation. Theoretical pursuit of the polar amplification problem 540 also points to the need for a more accurate understanding and better simulation of tropical 541 convection in climate models. The current level of understanding is apparently inadequate 542 (Stevens and Bony 2013). The following quote by Visser et al. (2003) from a paleoclimate 543 perspective underscores the need to further develop such a general-­‐circulation based 544 theory: “A new view of the importance of the tropics in controlling global climate change is 545 beginning to emerge, although the nature of the linkage between tropical and extra-­‐tropical 546 regions still needs to be resolved.” 547 548 Acknowledgements: This research was supported by Seoul National University, Republic of 549 Korea, and by NSF grant AGS-­‐1139970, USA. 550 551 552 553 554 26 References 555 556 557 Abbot, D.S. and E. Tziperman, 2008a: A high-­‐latitude convective cloud feedback and equable 558 climates. Q. J. R. Meteorol. Soc. 134, 165-­‐185. 559 560 Abbot, D.S. and E. Tziperman, 2008b: Sea ice, high-­‐latitude convection, and equable climates. 561 Geophys. Res. Lett. 35, L03702, doi:10.1029/2007GL032286. 562 563 Abbot, D.S., C.C. Walker, and E. Tziperman, 2009: Can a convective cloud feedback help to 564 eliminate winter and spring sea ice at high CO2 concentrations? J. Climate, 22, 5719-­‐5731. 565 566 Alexeev, V.A., P. L. Langen, and J. R. Bates, 2005: Polar amplification of surface warming on 567 an aquaplanet in “ghost forcing” experiments without sea ice feedbacks. Clim Dyn, 24, 655-­‐
568 666. 569 570 Bannon, P. R., 2012: Atmospheric available energy. J. Atmos. Sci., 69, 3745-­‐3762. 571 572 Barron, E.J., W.H. Peterson, D. Pollard and S.L. Thompson, 1993: Past climate and the role of 573 ocean heat transport: model simulations for the Cretaceous. Paleoceanography, 8, 785-­‐798. 574 27 575 Bralower, T. J., D. J. Thomas, J. C. Zachos, M. M. Hirschmann, U. Rohl, H. Sigudsson, H. E. 576 Thomas, and D. L. Whitney, 1997: High-­‐resolution records of late Paleocene thermal 577 maximum and circum-­‐Caribbean volcanism: Is there a causal link? Geology, 25, 963-­‐966. 578 579 Budyko, M. I., 1969: The effect of solar radiation variations on the climate of the earth. 580 Tellus, 21, 611-­‐619. 581 582 Budyko, M. I., and Izrael, Y. A., 1991: In Anthropogenic Climate Change, ed. M. I. Budyko, Y. A. 583 Izrael, pp. 277-­‐318. Tucson: Uni. Ariz. Press. 584 585 Caballero, R., and M. Huber 2010: spontaneous transition to superrotation in warm climates 586 simulated by CAM3, Geophy. Res. Lett., 37, DOI: 10.1029/2010GL043468. 587 588 Charney, J. G., and P. G. Drazin, 1961: propgation of planetary-­‐scale disturbances from the 589 lower into the upper atmosphere. J. Geophy. Res., 66, 83-­‐109. 590 591 Charney, J. G., M. E. Stern, 1962: On the instability of internal baroclinic jets in a rotating 592 atmosphere. J. Atmos. Sci. 19, 159-­‐172. 593 594 Chiang, John C. H., M. Biasutti, and D. S. Battisti, 2003: Sensitivity of the Atlantic Intertropical 595 Convergence Zone to Last Glacial Maximum boundary conditions, 596 Paleoceanography, 18(4). 10.1029/2003PA000916. 28 597 598 Cehelsky, P., and K. K. Tung, 1987: Theories of multiple equilibria and weather regimes—A 599 critical reexamination. Part II: Baroclinic two-­‐layer models. J. Atmos. Sci., 44, 3282–3303. 600 601 Clement, A. C., R. Seager, M. A. Cane, and S. E. Zebiak, 1996: An ocean dynamical thermostat. 602 J. Climate, 9, 2190–2196. 603 604 DeConto, R.M., E.C. Brady, J.C. Bergengren and W.W. Hay, 2000: Late Cretaceous climate, 605 vegetation and ocean interactions in Warm Climates in Earth History, B. R. Huber, K. G. 606 MacLeod, S. L. Wing, Eds., Cambrige Univ. Press, pp. 275-­‐296. 607 608 Douglas, R. G., and S. M. Savin, 1978: Oxygen isotopic evidence for the depth stratification of 609 Tertiary and Cretaceous planktic foraminifera. Marine Micropaleontology, 3(2), 175-­‐196. 610 611 Durack, P. J., and S. E. Wijffels 2010: Fifty-­‐year trends in global ocean salinities and their 612 relationship to broad-­‐scale warming. J. Climate, 23, 4342-­‐4362. 613 614 Farrell, B. F., 1990: Equable climate dynamics. J. Atmos. Sci. 47, 2986-­‐2995. 615 616 Flourney, M., S. B. Feldstein, S. Lee, and E. Clothiaux 2013: On the linkage between station 617 downward infrared radiation data, teleconnections, and tropical convection. In preparation. 618 29 619 Garfinkel, C. I., D. L. Hartmann, and F. Sassi, 2010: Tropical precursors of anomalous 620 Northern Hemisphere stratospheric polar vortices. J. Climate, 23, 3282-­‐3299. 621 622 Garfinkel, C. I., S. B. Feldstein, D. W. Waugh, C. Yoo, and S. Lee, 2012: Observed connection 623 between stratospheric sudden warmings and the Madden-­‐Julian Oscillation. Geophys. Res. 624 Lett., 39, http://dx.doi.org/10.1029/2012GL053144. 625 626 Graversen, R. G., 2006: Do changes in the midlatitude circulation have any impact on the 627 Arctic surface air temperature trend? J. Climate, 19, 5422-­‐5438. 628 629 Graversen, R. G., and M. Wang, 2009: Polar amplification in a coupled climate model with 630 locked albedo. Climate Dyn., 33, 629–643, doi:10.1007/s00382-­‐009-­‐0535-­‐6. 631 632 Hansen, J., A. Lacis, D. Rind, G. Russell, P. Stone, I. Fung, R. Ruedy, and J. Lerner, 1984: 633 Climate sensitivity: Analysis of feedback mechanisms. Climate Processes and Climate 634 Sensitivity, Geophys. Monogr., Vol. 29, Amer. Geophys. Union, 130–163 pp. 635 636 Held, I. M. and V. D. Larichev, 1996: Scaling theory for horizontally homogeneous, baro-­‐ 637 clinically unstable flow on a beta-­‐plane. J. Atmos. Sci., 53, 945-­‐952. 638 639 Held, I. M. and B. J. Soden, 2006: Robust response of the hydrological cycle to global 640 warming. J. Climate, 19, 5686–5699. 30 641 642 Hoffert, M. I., and C. Covey, 1992: Deriving global climate sensitivity from 643 palaeoclimate reconstructions. Nature, 360, 573-­‐576. 644 645 Honda, M., J. Inoue, and S. Yamane, 2009: Influence of low Arctic sea-­‐ice minima on 646 anomalously cold Eurasian winters, Geophys. Res. Lett., 36, L08707, 647 doi:10.1029/2008GL037079. 648 649 Hoskins, B. J., and D. Karoly, 1981: the steady linear response of a spherical atmosphere to 650 thermal and orographic forcing. J. Atmos. Sci., 38, 179-­‐1196. 651 652 Huber, B., K. G. MacLeod, and S. L. Wing, 2000: Warm Climates in Earth History. Cambridge 653 Press, 462 pp. 654 655 Huber, M. 2008. A hotter greenhouse? Science, 321, 353-­‐354. 656 657 Huang, H.-­‐P., K. M. Weickmann, and C. J. Hsu, 2001: Trend in atmospheric angular 658 momentum in a transient climate change simulation with greenhouse gas and aerosol 659 forcing. J. Climate, 14, 1525–1534. 660 661 Hwang, Y.-­‐T., and D. M. W. Frierson, 2010: Increasing atmo-­‐ spheric poleward energy 662 transport with global warming. Geophys. Res. Lett., 37, L24807, doi:10.1029/2010GL045440. 31 663 664 Johanneseen, O. M., L. Bengtsson, M. W. Miles, S. I. Kuzmina, V. A. Semenov, G. V. Alekseev, 665 A. P. Nagurnyi, V. P. Zakharov, L. P. Bobylev, L. H. Pettersson, K. Hasselmann, and H. P. Cattle, 666 2004: Arctic climate change: observed and modeled temperature and sea-­‐ice variability. 667 Tellus, 56A, 328-­‐341. 668 669 Kim, H. K. and S. Lee, 2001: Hadley cell dynamics in a primitive equation model. Part II: 670 Nonaxisymmetric flow. J. Atmos. Sci., 58, 2859-­‐2871. 671 672 Korty, R.L., K.A. Emanuel and J.R. Scott, 2008: Tropical cyclone-­‐induced upper-­‐ocean mixing 673 and climate: Application to equable climates. J. Climate, 21, 638-­‐654. 674 675 Koutavas, A., J. Lynch-­‐Stieglitz, T. M. Marchitto Jr., and J. P. Sachs, 2002: El Niño-­‐Like Pattern 676 in Ice Age Tropical Pacific Sea Surface Temperature Science 297, 226. DOI: 677 10.1126/science.1072376. 678 679 Kump, L.R. and D. Pollard, 2008: Amplification of Cretaceous warmth by biological cloud 680 feedbacks. Science, 320, 195. 681 682 L’Heureux, M. L., S. Lee, and B. Lyon, 2013: Recent multidecadal strengthening of the Walker 683 circulation across the tropical Pacific. Nature Climate Chage, DOI:10.1038/NCLIMATE1840. 684 32 685 Lea, D. W., D. K. Pak, and H. J. Spero, 2000: Climate impact of late quaternary equatorial 686 pacific sea surface temperature variations. Science, 289, 1719. DOI: 687 10.1126/science.289.5485.1719 688 689 Lee, S., 1999: Why are the climatological zonal mean winds easterly in the equatorial upper 690 troposphere? J. Atmos. Sci., 56, 1353–1363. 691 692 Lee, S. B. Feldstein, D. Pollard, and T. S. White, 2011a: Can planetary wave dynamics explain 693 equable climates? J. Climate, 24, 2391–2404. 694 695 Lee, S., T. Gong, N. Johnson, S. B. Feldstein, and D. Pollard, 2011b: 696 On the possible link between tropical convection and the Northern Hemisphere Arctic 697 surface air temperature change between 1958 and 2001. J. Climate, 24, 4350–4367. 698 699 Lee S. 2012. Testing of the Tropically Excited Arctic Warming Mechanism (TEAM) with 700 traditional El Niño and La Niña. Journal of Climate 25: 4015–4022, DOI: 10.1175/jcli-­‐d-­‐12-­‐
701 00055.1. 702 703 Li, L., A. P. Ingersoll, X. Jiang, D. Feldman, and Y. L. Yung, 2007: Lorenz energy cycle 704 of the global atmosphere based on reanalysis datasets. Geophys. Res. Lett., 34, L16813. 705 33 706 Lindzen, R. S., and Farrell, B. 1980: The role of polar regions in global climate, and the 707 parameterization of global heat transport. Mon. Weather Rev. 108:2064-­‐79. 708 709 Liu, J., J. A. Curry, H. Wang, M. Song, and R. M. Horton, 2012: Impact of declining Arctic sea 710 ice on winter snowfall. PNAS, 109, 4074-­‐4079, doi: 10.1073/pnas.1114910109. 711 712 Lorenz, E. N., 1955: Available potential energy and the maintenance of the general 713 circulation. Tellus, 7, 157-­‐167. 714 715 Lu, J., and M. Cai, 2010: Quantifying contributions to polar warming amplification in an 716 idealized coupled general circu-­‐ lation model. Climate Dyn., 34, 669–687. 717 718 Madden R. A., and P. R. Julian, 1971: Detection of a 40-­‐50 day oscillation in the zonal wind in 719 the tropical Pacific. J. Atmos. Sci., 28, 702-­‐708. 720 721 Madden R. A., and P. R. Julian, 1972: Description of global scale circulation cells in the 722 Tropics with 40-­‐50 day period. J. Atmos. Sci., 29, 1109-­‐1123. 723 Manabe, S., R. T. Wetherald, 1975: The effect of doubling the CO2 concentration on the 724 climate of a general circulation model. J. Atmos. Sci., 32, 3-­‐15. 725 726 Manabe, S., and R. J. Stouffer, 1980: Sensitivity of global climate model to an increase of CO2 727 concentration in the atmosphere. J. Geophy. Res., 85, 5529-­‐5554. 34 728 729 Masson-­‐Delmotte et al., 2006: Past and future polar amplification of climate change: climate 730 model intercomparisons and ice-­‐core constraints. Climate Dynamics, 26, 513-­‐529. 731 732 Matthews, A. J., B. J. Hoskins, and M. Masutani, 2004: The global response to tropical 733 heating in the Madden–Julian Oscillation during the northern winter. Quart. J. Roy. Met. 734 Soc., 130, 1991-­‐2011, doi: 10.1256/qj.02.123. 735 736 McKenna, M., 1980: Eocene paleolatitude, climate, and mammals of Ellesmere Island. 737 Palaeogeogr. Palaeoclimatol. Palaeoecol., 30, 349-­‐362. 738 739 Meehl, G.A., T.F. Stocker, W.D. Collins, P. Friedlingstein, A.T. Gaye, J.M. Gregory, A. Kitoh, R. 740 Knutti, J.M. Murphy, A. Noda, S.C.B. Raper, I.G. Watterson, A.J. Weaver and Z.-­‐C. Zhao, 2007: 741 Global Climate Projections. In: Climate Change 2007: The Physical Science Basis. Contribution 742 of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on 743 Climate Change [Solomon, S.,D. Qin, M. Manning, Z. Chen, M. Marquis, K.B. Averyt, M. 744 Tignor and H.L. Miller (eds.)]. Cambridge University Press, Cambridge, United Kingdom and 745 New York, NY, USA. 746 747 Miller, G. H., R. B. Alley, J. Brigham-­‐Grette, J. J. Fitzpatrick, L. Polyak, 748 M. C. Serreze, and J. W.C. White, 2010: Arctic amplification: can the past constrain the future? 749 Quaternary Science Reviews, doi:10.1016/j.quascirev.2010.02.008 35 750 751 Otto-­‐Bliesner, B.L. and G.R. Upchurch, 1997: Vegetation-­‐induced warming of high-­‐latitude 752 regions during the late Cretaceous period. Nature, 385, 804-­‐807. 753 754 Overland, J. E., and M. Wang, 2010: Large-­‐scale atmospheric circulation changes associated 755 with the recent loss of Arctic sea ice. Tellus, 62A, 1-­‐9. 756 757 Park, H.-­‐S., S. Lee, S.-­‐W. Son, Y. Kosaka, and S. B. Feldstein 2013: Rapid increase in Arctic 758 winter downward longwave radiation and sea ice melting. In preparation. 759 760 Pearson, P. N., P. W. Ditchfield, J. Singano, K. G. Harcourt-­‐Brown, C. J. Nicholas, R. K. Olsson, 761 N. J. Shackleton, and M. A. Hall, 2001: Warm tropical sea surface temperatures in the Late 762 Cretaceous and Eocene epochs. Nature, 413, 481-­‐487. 763 764 Pedlosky, J., 1964: The stability of currents in the atmosphere and ocean. Part I. 765 J. Atmos. Sci., 21, 201–219. 766 767 Peixoto, J. P., and A. H. Oort, 1992: Physics of Climate. American Institute of Physics, 520 pp. 768 769 Persson, P. O., 2012: Onset and end of the summer melt season over sea ice: thermal 770 structure and surface energy perspective from SHEBA. Clim Dyn 39:1349–1371. 771 DOI:10.1007/s00382-­‐011-­‐1196-­‐9 36 772 773 Petoukhov, V., and V. A. Semenov, 2010: A link between reduced Barents-­‐Kara sea ice and 774 cold winter extremes over northern continents, J. Geophys. Res., 115, 775 D21111, doi:10.1029/2009JD013568. 776 777 Pfeffer, R. L., 1981: Wave-­‐mean flow interactions in the atmosphere. J. Atmos. Sci., 38, 1340-­‐
778 1359. 779 780 Rickaby, R. E. M., and P. Halloran, 2005: Cool La Niña during the warmth of the Pliocene? 781 Science, 307(5717), 1948-­‐1952. DOI: 10.1126/science.1104666 782 783 Rigor, I. G., R. L. Colony, S. Martin, 2000: Variations in surface air temperature observations 784 in the Arctic, 1979-­‐97. J. Climate, 13, 896-­‐914. 785 786 Salmon, R., 1980: Baroclinic instability and geostrophic turbulence. Geophys. Astrophys. 787 Fluid Dyn., 15, 167-­‐211. 788 789 Saravanan, R., 1993: Equatorial superrotation and maintenance of the general circulation in 790 two-­‐level models. J. Atmos. Sci., 50, 1211–1227. 791 792 Sardeshmukh, P. D., and B. J. Hoskins, 1988: The Generation of Global Rotational Flow by 793 Steady Idealized Tropical Divergence. J. Atmos. Sci., 45, 1228-­‐1251. 37 794 795 Serreze, M. C., and R. G. Barry 2011: Processes and impacts of Arctic amplification: A 796 research synthesis. Quaternary Science Review (2011), doi: 10.1016/j.gloplacha.2011.03.004 797 798 Sewall, J.O. and L.C. Sloan, 2004: Arctic Ocean and reduced obliquity on early Paleogene 799 climate. Geology, 32, 477-­‐480. 800 801 Singarayer, J. S., J. L. Bamber, and P. J. Valdes, 2006: Twenty-­‐first-­‐century climate impacts 802 from a declining Arctic sea ice cover. J. Climate, 19, 11091125. 803 804 Sloan, L. C., J.C.G. Walker and T.C. Moore, 1995: The role of oceanic heat transport in early 805 Eocene climate. Paleoceanography, 10, 347-­‐356. 806 807 Sohn, B. J. and S.-­‐C. Park 2010: Strengthened tropical circulations in past three decades 808 inferred from water vapor transport. J. Geophys. Res. 115, D15112. 809 810 Spicer, R.A., A. Ahlberg, A.B. Herman, C.-­‐C. Hofmann, M. Raikevich, P. J. Valdes and P. J. 811 Markwick, 2008: The Late Cretaceous continental interior of Siberia: A challenge for climate 812 models. Earth Plan. Sci. Lett., 267, 228-­‐235. 813 814 Sriver, R.L. and M. Huber, 2007: Observational evidence for an ocean heat pump induced by 815 tropical cyclones. Nature, 447, 577-­‐580. 38 816 817 Stevens, B., and S. Bony, 2013: What are climate models missing? Science, 340, 1053; DOI: 818 10.1126/science.1237554. 819 820 Stone, P. H., 1978: Baroclinic adjustment. J. Atmos. Sci. 35:561-­‐71. 821 822 Stott, L., C. Poulsen, S. Lund, and R. Thunell 2002: Super ENSO and Global Climate 823 Oscillations at Millennial Time Scales Science 297, 222. DOI: 10.1126/science.1071627 824 825 Stroeve, J., M. C. Serreze, M. M. Holland, J. E. Kay, J. Malanik, and A. P. Barrett 2011: The 826 Arctic’s rapidly shrinking sea ice cover: a research synthesis, Climate Change. 827 DOI:10.1007/s10584-­‐011-­‐0101-­‐1 828 829 Tarduno, J. A., D. B. Brinkman, P. R. Renne, R. D. Cottrell, H. Scher, and P. Castillo, 1998: 830 Evidence for extreme climatic warmth from late Cretaceous Arctic vertebrates. Science, 282, 831 2241-­‐2244. 832 833 Vallis, G. K., 1988: Numerical studies of eddy transport properties in eddy-­‐resolving and 834 parameterized models. Quart. J. Roy. Meteor. Soc., 114, 183-­‐204. 835 836 Vecchi, G. A., and B. J. Soden, 2007: Global warming and the weakening of the tropical 837 circulation. J. Climate, 20, 4316–4340. 39 838 839 Visser, K., R. Thunell, and L. Stott, 2003: Magnitude and timing of temperature change in the 840 Indo-­‐Pacific warm pool during deglaciation. Nature, 421, 152-­‐155. 841 842 Walker, C. C. and T. Schneider, 2006: Eddy influences on Hadley circulations: Simulations 843 with an idealized GCM. Journal of the Atmospheric Sciences 63: 3333-­‐3350. 844 845 Walsh, J.E., W.L. Chapman, V.E. Romanovsky, J.H. Christensen, M. Stendel, 2008: Global 846 climate model performance over Alaska and Greenland. J. Climate, 21, 6156-­‐6174. 847 848 Wang X. and J. R. Key, 2005: Arctic surface, cloud, and radiation properties based on the 849 AVHRR polar pathfinder dataset. Part II: recent trends. J Clim 18:2575–2593 850 851 Winton M., 2006: Amplified Arctic climate change: what does surface albedo feedback have 852 to do with it. Geophys. Res. Lett., 33, doi: 10.1029/2005GL025244. 853 854 Wu, Y., M. Ting, R. Seager, H.-­‐P. Huang, and M. Cane, 2010: Changes in storm tracks and 855 energy transports in a warmer climate simulated by the GFDL CM2.1 model. Climate Dyn., 1–
856 20, doi:10.1007/s00382-­‐010-­‐0776-­‐4. 857 40 858 Yoo, C., S. Feldstein, and S. Lee, 2011: Impact of the Madden-­‐ Julian oscillation trend on the 859 Arctic amplification of surface air temperature during the 1979–2008 boreal winter. Geophys. 860 Res. Lett., 38, L24804, doi:10.1029/2011GL049881. 861 862 Yoo C, Lee S, Feldstein SB. 2012a. Mechanisms of extratropical surface air temperature 863 change in response to the Madden–Julian oscillation. J. Climate, 25, 5777–5790, DOI: 864 10.1175/jcli-­‐ d-­‐11-­‐00566.1. 865 866 Yoo C, Lee S, Feldstein S. 2012b. Arctic response to an MJO-­‐like tropical heating in an 867 idealized GCM. J. Atmos. Sci., 69, 2379–2393, DOI: 10.1175/JAS-­‐D-­‐11-­‐0261.1. 868 869 Zelinka, M. D., and D. L. Hartmann 2012: Climate Feedbacks and Their Implications for 870 Poleward Energy Flux Changes in a Warming Climate, J. Climate, 25, 608-­‐624. 871 872 873 874 875 876 877 878 879 41 880 881 882 883 884 885 886 887 888 889 890 891 892 © 1992 Nature Publishing Group
Figure 1: (a) Zonal-­‐mean surface temperature deviations (ΔT) from the present-­‐day climate divided by < ΔT >, where < > denotes a global average. The solid line is a least-­‐squares fit to the data displayed for a fourth-­‐order polynomial in the sine of latitude; NT(x) = ΔT(x)/<ΔT> = 0.36+3.21x4, where x = sin(latitude). (b) ΔT for the mid-­‐Cretaceous maximum warming and LGM cooling. Points are derived from paleodata and lines indicate the best-­‐fit NT(x) shown in (a) multiplied by the corresponding <ΔT>. [From Hoffert, M. I., and C. Covey, 1992]. 42 differed significantly from present. Barron et al. (1995) estimated
global average temperatures about 6 $ C warmer in the Cretaceous
than recently. Subsequent work suggests upward revision of tropical sea-surface temperatures by as much as a few degrees (Alley,
2003; Bice et al., 2006). The Cretaceous peak warmth seems to
have been somewhat higher than early Cenozoic values, or perhaps
similar (Zachos et al., 2001). Early Cenozoic summertime temperatures of w18 $ C in the Arctic Ocean and w17 $ C on adjacent Arctic
alies across
mplified by
Holocene,
rapid, and
same time
893 During the
894 ng modern
895 ombination
896 relative to
ure anom-
Observed Arctic Temperature Change (ºC)
hern HemiG may have
increased
mperatures
xpanded to
obal annual
20
y = 3.4x - 0.35
r2 = 0.99
MP
10
LIG
0
HTM
-10
le Pliocene
which had
-20
onnections
LGM
most recent
Matthiessen
ted terres-30
re available
-7
-5
-3
-1
1
3
5
obably no
Observed/modeled global temperature change (ºC)
e the Cana897 898 Matthews,
Fig. 4. Paleoclimate data quantify the magnitude of Arctic amplification. Shown are
paleoclimate estimates of Arctic summer temperature anomalies relative to recent, and
899 monarson,
the appropriate Northern Hemisphere or global summer temperature anomalies,
900 6). A global
together
with theireuncertainties,
for the
following:
the Last
Glacial Maximum
(LGM;
901 Figure 2: Paleodata stimates of Arctic summer surface temperature deviations from the al. (2008)
w20 ka ago), Holocene Thermal Maximum (HTM; w8 ka ago), Last Interglaciation
902 present-­‐day values (ordinate), their NH or global mean values (abscissa), and their malies were
(LIG; 130–125 ka ago) and middle Pliocene (w3.5 Ma ago). The trend line suggests that
903 uncertainties. The time periods shown are: the last Glacial Maximum (LGM), Holocene summer temperature changes are amplified 3–4 times in the Arctic.
malies of as
904 Thermal Maximum (HTM), Last Interglaciation (LIG), and middle Pliocene (MP). [From Miller 905 et al., (2010)] 906 can the past constrain the future?, Quaternary Science Reviews (2010),
amplification:
907 43 3896
JOURNAL OF CLIMATE
VOLUME 2
908 909 FIG. 6. Seven-year running mean time series of Northern Hemisphere zonal average surface
temperature
anomalies
910 Figure 3: Time series of the seven-­‐year running mean of (8C).
the anomalous zonally averaged 911 surface temperature (oC). [From Bekryaev et al., 2010)]. 912 atr as a measure of PA may lead to spring (Table 2, Fig. 5). The former can be explained b
999). Therefore,
913 unrealistic
estimates.
Figure 7c shows a time series of SDs local ice–albedo feedback mechanisms; however the la
914 derived from an ensemble of model-based ar and atr. This ter cannot. Indeed, this seasonal cycle of PA is consis
915 figure suggests that ar provides stable estimates of PA, tent with a large-scale winter and springtime ‘‘radiative
916 whereas the
917 wide
spread of atr precludes the use of atr as mechanism of PA and a year-round ‘‘advective’’ mech
meaningful
of PA. However, we expect that anism of PA, which is strongest in the cold season. Th
918 measure
with a warmer
the difference between atr and ar snow–albedo feedback mechanism may also be a plau
919 climate,
may diminish because an increasingly higher portion of sible explanation for the springtime PA. During the las
T9NPA(t) and T9NH(t) variance will be yielded by trends.
15–20 years, PA during the cold season was very stron
The steadiness of model-based ar (Fig. 7d) suggests that (Figs. 3 and 8). For example, in 2006–08 high-latitud
he relationship between NH and NPA SATs is preserved winter–spring SATs were exceptionally high (Fig. 4
egardless of whether the preindustrial period or the Unfortunately, observational SAT data preclude us from
ast decades of the strongest anthropogenic warming are separating the contribution of ‘‘advective’’ and ‘‘radia
onsidered. Polar amplification may be described as an tive’’ mechanisms into the large-scale PA.
ntrinsic mode of climate variability (Alexeev et al. 2005;
To ensure that our estimates of large-scale PA are no
Langen and Alexeev 2007); therefore, human-induced ‘‘contaminated’’ by effects of local ice–albedo feedback
limate change will be projected on this mode without we repeated our calculations of ar presented in Table 2 b
ltering its spatial structure, in agreement with the con- using NPA records from meteorological stations locate
ept of nonlinear climate dynamics perspective proposed at least 250 km away from the coast (Table 3). In genera
by Palmer (1999). In that case, climate change and its the differences between these two calculations are smal
ariations at various time scales will produce similar es- within 15%. However, there is a substantial, almos
twofold, increase of PA during the cold season. Tha
imates of PA.
Large-scale PA derived from observational data is may be a manifestation of a positive feedback mecha
vident from Table 2 with annual 1875–2008 ar 5 1.62 6 nism among the SAT, available precipitable water, an
.10, close to the 1.56 estimated from ensemble-average 44 incoming longwave radiation, which is strongest in re
PA derived from a suite of 42 CMIP model runs. There gions with the lowest SATs. The inland meteorologica
s a seasonal cycle with maximum PA in autumn and stations have the lowest winter and spring SATs. Sinc
(a)$Reference$Climate$$$
Poleward energy flux
deficit
solar
terrestrial
surplus
deficit
Poleward energy flux
920 (b)$Equable$Climate$$$
Poleward energy flux
Larger deficit
solar
terrestrial
surplus
Larger deficit
921 922 923 924 925 926 927 928 929 Poleward energy flux
Figure 4: Schematic diagram of top of the atmosphere (TOA) net radiation and poleward energy flux. According to Fig. 1, under equable climates, warming is much greater in high latitudes. This implies that the increase in outgoing long wave radiation (OLR) is greater in the high latitudes, hence strengthening the equator-­‐to-­‐pole gradient in TOA net radiation. This implies that the poleward energy flux must be stronger under equable climates. Note that this balance between the TOA radiation gradient and the poleward energy flux, on its own, cannot be used to infer a causal relationship. 45 930 931 932 933 height$
(a)$Poleward$eddy$heat$flux$
(2)$eastward$accelera.on$at$the$wave$
source;$if$strong$enough,$generates$$
equatorial$superrota.on$
(4)$Storm$track$
eddies$may$$
respond,$and$$
transport$heat$
(1)$Zonally$
farther$poleward$
localized$
convec.ve$
hea.ng$$
(3)$Poleward$and$upward$
propaga.ng$waves$
transport$heat$poleward$
934 935 936 937 equator$
pole$
pole%
(b)%Adiaba3c%warming%
(5)%Par3al%wave%absorp3on%serves%as%a%
wave%sink,%causing%westward%
accelera3on%
X
height%
∂u
<0
∂z
(6)%Thermal%wind%balance%requires%
adiaba3c%warming%(sinking%mo3on)%
poleward%of%the%wave%sink%
938 939 940 941 equator%
€
46 942 943 944 945 (c)%Downward%IR%
height%
(8)%Moisture%condenses,%
clouds%form,%and%emit%
downward%IR%
(7)%Poleward%heat%flux%is%
accompanied%by%
moisture%flux%%
pole%
equator%
946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 Figure 5: Schematic diagram of the TEAM mechanism. The storm track eddy response in step (4) and the cloud cover increase in step (8) are found to accompany the forced tapping (of ZAPE) mechanism (Lee et al. 2011a). Step (7) is shown by Yoo et al. (2012a, 2012b), and (8) by Yoo et al. (2012b). It is worth noting that the equatorial superrotation can be a byproduct of the TEAM mechanism. 47 964 (a)$
965 (b)$
966 967 968 969 970 971 972 dam$
Figure 6: January difference in (a) surface temperature and (b) 250-­‐hPa geopotential height and wind fields between a run with 150 W/m2 contrast (between the red and blue regions) and a run with no heating contrast. The model is a coupled atmosphere-­‐mixed-­‐layer ocean GCM. The CO2 level in this model is 4 × PAL (Preindustrial Atmospheric Level, 1 PAL = 280 ppmv). [Adopted from Lee et al. 2011a.] 48 (a) 4 X PAL CO2
30
20
0
75
120
150
180
225
15
10
5
0
977 978 979 980 981 982 983 984 985 986 987 10
0
75
120
150
180
225
5
0
o
Temperature (oC)
25
(b) 4 X PAL CO2
Temperature ( C)
973 974 975 976 −5
−10
−15
10
20
30
40
Latitude (degrees)
−20
50
60
70
80
Latitude (degrees)
90
Figure 7: January zonal mean surface air temperature simulated by a coupled atmosphere-­‐
mixed-­‐layer ocean GCM. In the legend, the values indicate the imposed heating difference between the red and blue boxes (W/m2) shown in Figure 6a. [Adopted from Lee et al. 2011a.] 49 (a)!
Warmer!air!temperature!in!all!!!!!!!!!!
!seasons!(GHG!forcing)!
Enhanced!!
ice1albedo!
feedback!
Earlier!development!of!
open!water!
Thinner!spring!ice!
More!open!water!
In!September!
!!!Warmer!autumn!
!!!!!temperatures!
988 (b)!
Warmer!air!temperature!in!all!!!!!!!!!!
!seasons!(GHG!forcing)!
Enhanced!!
ice1albedo!
feedback!
Synop>c!weather!
events!can!more!
readily!melt!ice!
?!
Thinner!spring!ice!
More!open!water!
In!September!
989 990 991 992 993 994 995 996 997 ?!
!!!Warmer!autumn!
!!!!!temperatures!!
Winter!&!spring!
warming!by!TEAM!
mechanism!
Figure 8: (a) Adopted from Stroeve et al. (2011), and shown here to be compared with (b). (b) Supported by the sea-­‐ice-­‐area autocorrelation shown in Fig. 9, the physical processes described by the TEAM mechanism suggest that winter Arctic conditions may influence summer Arctic sea ice. The question marks indicate that the causal relationship may not be as solid as previously suggested (e.g., by Stroeve et al. 2011) for the present warming. These processes may become more important as sea ice continues to melt. 50 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 Figure 9: Lagged autocorrelation of monthly mean Arctic sea-­‐ice area from 1979-­‐2010. The data is from National Snow and Ice Data Center. The abscissa is the reference month and the ordinate is the corresponding lag month. For instance, (4,-­‐7) is for correlation between April and the previous September sea-­‐ice area. Except for the dotted area, all values are statistically significant (p < 0.05). To aid visualization, the abscissa shows double of the twelve-­‐month period. 51