By analyzing El Niño and La Niña composites with 40-yr European Centre for Medium-Range Weather Forecasts (ECMWF) Re-Analysis (ERA-40) data, evidence is presented here that the surface air temperature of the Arctic winter (December–February) is anomalously warm during La Niña and cold during El Niño. Surface and top-of-the-atmosphere energy fluxes were used to calculate the composite zonal-mean poleward moist static energy transport. The result shows that the La Niña warming in the Arctic is associated with an increased poleward energy transport in the extratropics. The opposite characteristics are found for El Niño. Because the total tropical convective heating is more localized during La Niña than El Niño, these findings suggest that the Arctic surface air temperature anomalies associated with ENSO may be attributed to the tropically excited Arctic warming mechanism (TEAM). In the tropics, consistent with previous studies, the anomalous poleward energy transport is positive during El Niño and negative during La Niña. Given the debate over whether a warmer world would take on more El Niño–like or La Niña–like characteristics, the findings of this study underscore the need for further investigation of tropical influence on polar climate.
Paleoclimate reconstructions and model simulations with enhanced greenhouse gas (GHG) forcings show that the surface meridional temperature gradient was weaker (stronger) during warm (cold) periods of the Earth’s history (e.g., Budyko and Izrael 1991, 277–318; Hoffert and Covey 1992). To maintain a weak meridional temperature gradient, an enhanced poleward heat transport is required, yet this cannot be accomplished by an increase in atmospheric baroclinic eddy heat flux because this flux is proportional to the zonal-mean temperature gradient. For this reason, proposed theories of equable climates involve increased poleward heat transport by other processes such as a global-scale Hadley circulation (Farrell 1990) or an enhanced ocean circulation (Barron et al. 1993; Korty et al. 2008). There are also theories of polar amplification that do not involve changes in poleward heat transport (e.g., Sewall and Sloan 2004; Abbot and Tziperman 2008a,b), the most prominent being the surface albedo feedback (SAF) mechanism (Budyko 1969; Sellers 1969). However, if the polar warming enhances outgoing longwave radiation in that region, the meridional gradient in the top-of-the-atmosphere (TOA) net radiation would increase, and thus a theory for meridional heat transport is still required.
Recently, an alternative mechanism was proposed that states that enhanced and localized tropical convection can contribute toward winter high-latitude warming, through the excitation of poleward-propagating Rossby waves (Lee et al. 2011a, hereafter L11a; Lee et al. 2011b, hereafter L11b). In this tropically excited Arctic warming mechanism (TEAM), the poleward-propagating Rossby waves can warm the Arctic through adiabatic warming, an enhanced poleward stationary eddy heat transport, and downward infrared (IR) radiation. The TEAM of L11a and L11b was inspired in part by the findings of Saravanan (1993), who showed with an atmospheric general circulation model that a “normal,” equatorially retrograding state undergoes a transition to an equatorially superrotating state if an imposed zonal-wavenumber-2 heating in the tropics exceeds a threshold value. In spite of the identical meridional temperature gradient in the model’s thermal forcing, his Fig. 1e shows that the midlatitude eddy heat flux in the superrotating state is only about half that of the normal state. Although this result is based on a dry two-layer model where the equatorial superrotation takes on an exaggerated form compared with multilayer model, it implies that there are other atmospheric processes that can significantly warm high latitudes and/or cool the tropics in that model.
There are two prominent internal processes that can stir the tropical atmosphere through convective heating: the Madden–Julian oscillation (MJO; Madden and Julian 1971) and El Niño–Southern Oscillation (ENSO). Consistent with the TEAM, Yoo et al. (2011, 2012b) found that MJO phase-5 convection (enhanced convection over the western tropical Pacific warm pool) is associated with an eastward acceleration in the equatorial upper troposphere and with a warming of the Arctic during the winter. Similarly, one can ask if the Arctic winter would be warmer during La Niña than during El Niño because the tropical convection is more localized for La Niña.1 Accordingly, in this study we address the following questions: Are Arctic surface temperatures higher for La Niña than for El Niño? If so, is the La Niña Arctic warming accompanied by a greater poleward energy flux into the Arctic?
We use the monthly-mean global 40-yr European Centre for Medium-Range Weather Forecasts (ECMWF) Re-Analysis (ERA-40) for the Northern Hemisphere (NH) winter months [December–February (DJF)]. The study focuses on the winter hemisphere (the Arctic rather than Antarctica for December through February) for two reasons: first, poleward-propagating Rossby waves in the winter hemisphere are much stronger than those during the summer hemisphere; second, polar regions are persistently exposed to the sun during the summer and, as a result, local shortwave radiative processes may play a much more important role than dynamically driven processes. In fact, Yoo et al. (2012a) found that the high-latitude MJO-driven surface air temperature (SAT) trend is negligible in the summer hemisphere for both the Arctic and Antarctic.
The ENSO events were identified with the oceanic Niño index (http://www.cpc.ncep.noaa.gov/products/analysis_monitoring/ensostuff/ensoyears.shtml). During 1957–2001, there are 13 El Niño events (1957/58, 1963/64, 1965/66, 1968/69, 1969/70, 1972/73, 1976/77, 1977/78, 1986/87, 1987/88, 1991/92, 1994/95, and 1997/98) and 14 La Niña events (1962/63, 1964/65, 1967/68, 1970/71, 1971/72, 1973/74, 1974/75, 1975/76, 1984/85, 1988/89, 1995/96, 1998/99, 1999/2000, and 2000/01). To obtain composites, we first calculated individual DJF means for each grid point, followed by the removal of their linear trends over the 1957–2001 period. Throughout this paper, an anomaly is defined as the deviation from the detrended DJF climatology of the ERA-40 data.
Because the ERA-40 data have a warm bias in the Arctic temperature trend, whereas such a bias is absent the ERA-Interim (Screen and Simmonds 2010), we also analyzed the ERA-Interim (1979–2010) reanalysis and compared the composite SAT (specifically the 2-m temperature) for the overlapping periods. The SAT composites over the Arctic are almost indistinguishable (not shown), with the pattern correlations between the two composites, for a domain confined to poleward of 60°N, being 0.97 and 0.98 for El Niño and La Niña, respectively. This result suggests that the main conclusion of this study is unlikely to be dependent on the data quality improvement from ERA-40 to ERA-Interim.
For all composites presented in this study, statistical significance is evaluated by employing a Monte Carlo test; for each grid point of a given composite, the probability distribution is constructed with 1000 repetitions of the same calculation using N (out of the 45) randomly selected DJF-mean values, where N is 13 for El Niño and 14 for La Niña.
3. SAT and poleward energy transport
The composite anomalous SATs (Fig. 1) conform to the canonical characteristics of the tropics and subtropics being warm during El Niño and vice versa during La Niña, yet they are also consistent with the TEAM, with a cold (warm) Arctic Ocean during El Niño (La Niña). Not surprisingly, the grid points that are statistically significant above the 95% confidence level are mostly confined to the tropics. However, there are also Arctic (defined as the region north of the Arctic Circle, 66.56°N) grid points that are still significant above the 95% confidence level: the El Niño cooling over the East Siberian Sea and the La Niña warming centered at the Kara Sea. Similar ENSO Arctic SAT anomalies are also present in an atmospheric general circulation model (AGCM) simulation forced with an observed SST field (Sassi et al. 2004).
Admittedly, these statistically significant grid points cover only a small fraction of the Arctic. Therefore, one may ask whether the TEAM can account for the observed interdecadal Arctic warming trend. However, it is important to stress that the ENSO is only one form of tropical variability and that the TEAM is one of several processes that may contribute toward Arctic warming. In addition, because both the ENSO and Arctic Oscillation (AO) are linked to the stratospheric quasi-biennial oscillation (QBO) [e.g., see Gray et al. (1992) for the ENSO–QBO relationship and Baldwin and Dunkerton (1999) for the AO–QBO linkage], it is to be expected that the Arctic SAT signal associated with the ENSO may include an AO contribution. Therefore, the impact of the TEAM may be weaker on the ENSO time scale than it is on other time scales.
According to the TEAM, downward IR radiation is an important contributor to the warming. Indeed, for the Arctic winter, where the incident solar radiative flux is essentially zero, Fig. 2b shows that an increase (a decrease) in anomalous downward IR radiation, which leads to a loss (gain) of atmospheric energy to (from) the surface, tends to occur where the SAT anomaly (Fig. 1) is positive (negative), although there are only a small number of statistically significant grid points over the Arctic Ocean. Over the same region where the amplitude of the SAT anomaly is large, the surface sensible and latent heat flux anomalies (FSH + FLH; Figs. 2c,d) are negligible.
Employing an anomalous atmospheric energy balance equation for the composite La Niña and El Niño, the above surface energy fluxes can be combined with the TOA energy fluxes (Figs. 2e,f) to estimate the anomalous poleward energy flux,
where E is moist static energy per unit area, υ is the meridional wind, is the net anomalous TOA downward energy flux, is the net anomalous upward surface energy flux, a is the radius of the earth, λ is the longitude, and φ is the latitude. Specifically, , where is absorbed TOA shortwave radiation and is outgoing TOA longwave radiation; , where and are downward surface shortwave and longwave radiative fluxes, respectively. The square bracket denotes a zonal integral such that . This indirect method of calculating the poleward energy flux was employed previously, for example, by Zhang and Rossow (1997), Trenberth et al. (2002), and Mayer and Haimberger (2012). In particular, the latter two studies examined variations in the flux associated with El Niño. The energy flux into the Arctic was not the focus of these studies, and none of these studies show the zonal-mean poleward energy flux poleward of 60°N. However, as will be shown below, at least in the tropics and subtropics, our estimate is consistent with these two previous studies.
Because the La Niña and El Niño composites of the left-hand side (lhs) of (1) are close to zero, following Trenberth et al. (2002), the northward energy flux can be estimated by integrating (1) with respect to latitude; that is,
where we assume that the moist static energy flux is zero at the South Pole. Figure 3a shows that the global mean of is essentially zero for La Niña (the thin and thick blue lines overlap), but it is positive for El Niño (the thin red line is shifted upward of the thick red line). This positive global-mean value is consistent with the finding of Trenberth et al. (2002) that the El Niño increases the global-mean surface temperature. Because the meridional energy flux should be zero at both Poles, we subtract the global mean of from the right-hand side (rhs) of (2) to estimate Eυ(φ): Upon integrating (1) from the South Pole to the North Pole, the first term on the rhs is identically zero, and we have
If we assume that the lhs of (4) is zero (i.e., the local rate of change in [E] is uniform or by simply assuming that the time tendency terms are negligible), by integrating (4) with respect to φ from −π/2 (South Pole), we arrive at (2) with subtracted from the rhs.
The result, shown in Fig. 3b, indicates that, in the deep tropics, the anomalous poleward energy flux is positive for El Niño and negative for La Niña. This is consistent with the findings of Trenberth et al. (2002) and Mayer and Haimberger (2012). In the extratropics, the La Niña anomalous poleward energy flux is in a poleward direction over most of the NH and for El Niño this flux is in an equatorward direction. Given that there are nonnegligible errors in the surface sensible and latent heat fluxes (Mayer and Haimberger 2012) as well as the TOA radiation fluxes (Trenberth and Fasullo 2010), this finding should be viewed as tentative. Mayer and Haimberger (2012) also showed that the accuracy of these estimates can be further improved by taking into account mass imbalance.
In spite of the various sources of error in our estimates, the extratropical poleward energy transport, as revealed by Fig. 3b, is consistent with the anomalous adiabatic warming/cooling and anomalous storm tracks associated with the ENSO; Seager et al. (2003) showed that the El Niño mean meridional circulation (MMC) adiabatically warms the subtropics and cools the midlatitudes in both hemispheres. Conversely, their results imply that during La Niña the MMC cools the subtropics and warms the midlatitudes. Seager et al. (2010) showed that Pacific storm track is shifted poleward during La Niña, with positive anomalous La Niña–El Niño values poleward of 45°N.
The zonal-mean zonal wind and MMC anomalies (indicated by the omega field) shown in Fig. 4 suggest that the eddies play a more important role than the overturning MMC in driving subtropical jet variability. This is because the largest zonal wind anomalies in the upper troposphere occur at ≈30° in both hemispheres, whereas a strengthened (weakened) Hadley circulation during El Niño (La Niña) terminates at ≈15° (see the omega field in Fig. 4), indicating that the maxima in the anomalous angular momentum conserving wind occurs at ≈15°, not at ≈30°. This implies that the subtropical zonal wind anomalies are driven by eddies. Because tropically excited Rossby waves can decelerate the subtropical zonal wind (Lee 1999; L11a), Fig. 4 suggests that the positive zonal wind anomalies in the subtropics during El Niño may be caused by anomalously weak poleward-propagating Rossby waves and the opposite during La Niña. This raises the possibility that Rossby waves excited by tropical convection may play a more significant role in shaping the ENSO-related zonal-mean flow anomalies than previously thought.
4. Discussion and concluding remarks
In this study, we present evidence for the Arctic winter that the surface air is anomalously warm during La Niña and cold during El Niño. It is found that the La Niña (El Niño) temperature change is associated with an increase (decrease) in the poleward moist static energy transport by the atmosphere, which was inferred from the TOA and SFC energy fluxes. Although the quantitative value of the energy transport should be taken with caution, the qualitative features of the implied poleward energy transport are consistent with the previously documented circulation changes associated ENSO (Seager et al. 2003, 2010). Because tropical convective heating is more localized during La Niña than El Niño, together with the evidence that poleward-propagating Rossby waves excited by tropical convection are stronger during La Niña (Lee 1999), our findings suggest that the TEAM process may contribute to the Arctic winter SAT increase (decrease) during La Niña (El Niño).
Given the above findings, it would be helpful to examine how the moist static energy flux is partitioned among eddy fluxes (stationary and transient) and MMC and among the four components of the moist static energy and how they are related to the circulation anomalies (e.g., teleconnections and storm tracks). Also, given that some El Niño events have highest SST anomalies in the central Pacific, close to the western warm pool (Larkin and Harrison 2005a,b; Ashok et al. 2007; Weng et al. 2007; Kao and Yu 2009; Kug et al. 2009), separating out these warm pool El Niño events from the composite may also help to contrast the difference between El Niño and La Niña in the Arctic. It is also important to keep in mind that there are processes other than the TEAM that influence the ENSO SAT anomalies over the Arctic and that these processes are not independent of each other. For example, as was discussed earlier, the stratospheric QBO can influence the Arctic SAT through its impact on the AO. However, it is also possible that the AO itself is influenced by tropical SST (Hoerling et al. 2001).
What do these results imply about the climate response to GHG forcing? There is circumstantial evidence that the TEAM process may intensify as climate warms. Lee (1999) showed with an aquaplanet GCM that the simulated MJO intensifies in response to a uniformly raised SST and that this results in a state of equatorial superrotation; Huang et al. (2001) found with a coupled GCM that the equatorial upper-tropospheric zonal wind changes into a superrotating state in response to a CO2 tripling. Because the tropics and subtropics are warmer during El Niño, it is sometimes likened to a GHG-driven warming2 (Knutson and Manabe 1995; Meehl and Washington 1996). On the other hand, as was shown here, in terms of winter Arctic warming, La Niña is more in line with GHG-driven warming. One possible conclusion is that it may be an enhancement in the zonal localization of tropical convective heating, rather than the overall tropical heating itself, that contributes to the winter Arctic warming.3 Given the evidence that the TEAM operates both in models and in the atmosphere (L11a; L11b; Ding et al. 2011; Yoo et al. 2011 ,2012a,b), it would be worthwhile to investigate whether this mechanism also operates in the IPCC Fifth Assessment Report models and whether its presence or absence (which depends on how faithfully the models simulate tropical convective heating) can help to explain intermodel variability in polar amplification.
This work was motivated by a conversation with Dr. Isaac Held. The author also thanks Dr. Steven Feldstein for comments on the manuscript. This work was supported by National Science Foundation Grant AGS-1139970.
As was suggested by Lu and Cai (2010) and Zelinka and Hartmann (2012), an increase in net tropical heating may warm high latitudes through increased baroclinic wave eddy heat flux in the upper troposphere. In contrast to the TEAM, however, this mechanism would predict a warmer Arctic during El Niño.
In terms of changes in the zonal gradient of the tropical Pacific SST, it is uncertain as to whether GHG forcing would increase (La Niña–like) or decrease (El Niño–like) the SST gradient. Vecchi et al. (2008) provide a summary of this debate.
It appears that enhancement and localization of tropical convective precipitation within the warm pool region is sometimes objected to on the grounds that the Walker circulation is predicted to weaken (Betts and Ridgway 1989; Held and Soden 2006). A common argument against the TEAM contributing to the Arctic warming trend is that a weakening of the Walker circulation should lead to a reduction in the west–east contrast in the tropical Pacific precipitation. However, if the dry static stability also increases, the Walker circulation may weaken while the precipitation–evaporation pattern intensifies.