Atmospheric energy transport to the Arctic 1979–2012

The Arctic troposphere has warmed faster than the global average over the last several decades. It was suggested that atmospheric northward energy transport (ANET) into the Arctic had contributed to tropospheric warming in the Arctic. Here we calculate ANET based on the NCEP/NCAR reanalysis data from 1979 to 2012. During this period the zonally integrated energy flux into the Arctic has decreased rather than increased in all seasons. However, the trends are statistically insignificant except for the winter and annual mean fluxes. For the winter season, the transient eddy flux of energy increases over Greenland and the Greenland Sea and decreases over west-central Siberia (WCS). Trends in meridional wind variance and vorticity also indicate increasing transient eddy activity over Northern Canada, the Greenland Sea and the Norwegian Sea and decreasing activity over WCS. Inter-winter variations in local vorticity over the WCS are significantly anti-correlated with the Arctic climate.


Introduction
Arctic amplification (AA) of global warming has been observed in both the upper and lower troposphere in the last few decades (e.g., Graversen et al., 2008;Alexeev et al., 2012;Screen et al., 2012). Chung et al. (2013) evaluated the warming in the Arctic using four different reanalysis data sets and radio-soundings for the period 1979Á2011. Significant warming trends are found for all seasons in all data sets, and two times greater rates of warming are found in the lower troposphere than in the upper troposphere. The annual mean surface temperature poleward of 708N increased by a range of 1.7590.29 K to 2.7190.28 K over 33 years, 4Á6 times larger than the global average surface air temperature increase (Arguez et al., 2013), and the warming in winter months (DecemberÁJanuaryÁFebruary, hereafter DJF) was about twice as large as in summer months (JuneÁ JulyÁAugust). With four-times present CO 2 in the atmosphere, contemporary global climate models predict a surface warming of 4.3 K in the tropics (308SÁ308N) and 11.2 K in the Arctic (608NÁ908N), with a range of 3Á6 K and 5Á16 K, respectively (Pithan and Mauritsen, 2014). AA of climate change has been observed in geological records and is now recognised as an inherent characteristic of the global climate system (Serreze and Barry, 2011). Multiple feedback effects can contribute to AA, including albedo, lapse rate and water vapour (Cai, 2005;Winton, 2006;Taylor et al., 2013;Graversen et al., 2014;Pithan and Mauritsen, 2014). Additional warming may result from changes in arctic vegetation (Swann et al., 2010) and black carbon in snow (Flanner et al., 2007).
The arctic warming over the past decade has been accompanied by a rapid decline in sea ice concentrations in the polar region (Comiso et al., 2008;Comiso, 2012). Changes in sea ice extent impact heat fluxes between the ocean and the atmosphere (Smedrud et al., 2013), and may have contributed to the Arctic surface warming (Serreze et al., 2009;Screen and Simmonds, 2010). In model simulations, local sea ice and sea surface temperature (SST) trends explain a large portion of the observed Arctic nearsurface warming, whereas remote SST changes explain the majority of observed warming aloft (Screen et al., 2012). Over longer time scales, the multidecadal oscillation of the Atlantic meridional overturning circulation (AMOC) may modulate Arctic sea ice and surface air temperature (SAT) (Chylek et al., 2009;Mahajan et al., 2011;Peings and Magnusdottir, 2014).
The vertical structure of the recent arctic warming as represented in observed data and reanalysis products has gained much deserved attention (Graversen et al., 2008;Alexeev et al., 2012;Screen et al., 2012;Chung et al., 2013). The warming aloft is not attributable to shortwave and longwave (LW) feedbacks, nor to changes in sea ice and SST *Corresponding author. email: Songmiao.Fan@noaa.gov Tellus A 2015. # 2015 S.-M. Fan et al. This is an Open Access article distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), allowing third parties to copy and redistribute the material in any medium or format and to remix, transform, and build upon the material for any purpose, even commercially, provided the original work is properly cited and states its license. within the Arctic. Graversen et al. (2008) proposed atmospheric northward energy transport (ANET) as an important mechanism for the recent arctic warming. Mid-tropospheric temperatures in the Arctic are sensitive to advection of energy into the Arctic: the 500 hPa temperature field in the Arctic was found to be positively correlated with ANET across 608N with a lag of 5 days (Graversen et al., 2008). The correlation may be robust on the synoptic time scale, but, on longer time scales, a reduced zonal-mean meridional temperature gradient may cause a decrease of energy transport (Serreze and Barry, 2005).
Other studies have searched for changes in large-scale atmospheric circulation patterns as potential causes. Remote SST patterns might affect arctic meteorological conditions through teleconnections (Screen et al., 2012;Perlwitz et al., 2015). However, details of the dynamical processes are not known to date. In a recent study based on the ERA-Interim reanalysis product, Woods et al. (2013) found that the interannual variability of intense moisture intrusions across 708N is strongly correlated with variability in wintermean surface downward LW radiation and skin temperature averaged over the Arctic. The intense intrusions are caused by large-scale blocking patterns which deflect mid-latitude cyclones and their associated moisture poleward. The above studies thus emphasise the importance of understanding the variability in large-scale circulation and its impacts on energy transport into the Arctic, the surface energy budget and top-of-atmosphere (TOA) net radiative flux.
In this study, we first extend the analysis of Graversen et al. (2008) based on ERA-40 reanalysis data to NCEP/ NCAR reanalysis data. We examine energy transport by planetary and transient waves through heightÁlongitude cross-sections and by the zonal mean circulation (ZMC). Section 2 describes the data and methods. Section 3 presents the energy fluxes and trends. Section 4 discusses inter-annual variations in winter. Section 5 summarises the results and conclusions.

Data and analysis method
The total energy per unit mass of air is given by E 0c v T' gz'Lq'½(u 2 'v 2 ), where c v is heat capacity, T is temperature, g is gravity, z is geopotential height, L is latent heat, q is specific humidity, u is zonal wind and v is meridional wind. The energy balance in a polar cap is given by Peixoto and Oort (1992) (Eq. 13.38 with the over bars for time averaging neglected): where m is mass, c p is heat capacity at constant pressure, v is meridional wind, x is zonal distance, p is pressure (p 0 p s s at a sigma-level, p s 0surface pressure), F TA and F TB are downward energy fluxes at top and bottom of the atmosphere, respectively. We calculate mass-weighted surface-TOA and surface-400 hPa northward fluxes of internal energy (denoted by vT), potential energy (vz) and latent heat (vq) using the NCEP/ NCAR reanalysis data from 1979 to 2012 (Kalnay et al., 1996). The reanalysis data is archived at 6-hourly intervals, 28 vertical sigma-levels, 192 longitude bands and 94 latitude bands. The kinetic energy flux is small and is neglected. We decompose the fluxes into contributions by stationary waves (denoted by v*q*, v*T* and v*z*) and transient eddies (v'q', v'T' and v'z') (Peixoto and Oort, 1992). We define transient eddies as departures from monthly mean quantities in each grid, and stationary waves as departures from zonal mean averages for each vertical level and in each month. Vertically averaged zonal mean v is subtracted from the mean v in each layer to ensure a zero net mass flux through a ''wall'' encircling the Arctic and to eliminate spurious results (Trenberth, 1991;Graversen, 2006). We also calculate monthly, mass-weighted variances of meridional wind (v'v'), mean positive relative vorticity (vort) and the time fraction for which vort !1.0 ) 10 (5 s (1 as measures of cyclone activity (Sorteberg and Walsh, 2008). Analyses are performed for twelve 308-wide longitudinal sectors (Fig. 1). The time series for each crosssection are analysed for interannual variability and linear trends from 1979 to 2012. Unless otherwise noted all analyses are performed for northern winter (DJF).

Energy fluxes by transient eddy and trends
In this section, we present an analysis of poleward energy fluxes. Figure 2 shows the ANET fluxes at 65.78N and 71.48N, which are divided by the encircled surface areas. The monthly mean fluxes (Fig. 2a) at 71.48N are in good agreement with a previous calculation at 708N based on a 10-year data set (Nakamura and Oort, 1988, as tabulated in Serreze and Barry, 2005). The annual cycle of ANET shows a broad peak from November to March and a minimum from May to July/August. The annual and seasonal mean ANET show large interannual variations and downward trends (Fig. 2bÁf). The linear trends from 1979 to 2012 are shown in Table 1 for ANET at 65.78N. The annual and winter trends are significant at 2 standard deviations (s.d.), while the trends in other seasons are insignificant. In comparison, calculations of ANET based on ERA-40 reanalysis data for 1979Á2001 show a downward trend in winter and an upward trend in summer (Graversen, 2006;Graversen et al., 2008). The trend in ANET at 608N explains only a fraction (e.g., 20Á30% near 700 hPa) of the temperature trends in summer (Graversen et al., 2008).
A partitioning of the ANET in winter is shown in Table 2. Transport of sensible heat is much larger than latent heat and together they tend to warm the Arctic. Transport of geopotential energy is dominated by the ZMC which tends to cool the Arctic. Transport by stationary waves is comparable to that by transient eddies. The total flux at 65.78N (2516 TW) is larger than that at 71.48N (1545 TW) due to a net radiative cooling in between the latitudes. Figure 3 shows the averages and linear trends of the energy flux terms by transient eddies through the cross-sections from the surface to Â400 hPa in winters from 1979 to 2012, and the same for the variances of v and for the averages of positive vorticity. The effect of transient baroclinic waves on energy transport are most pronounced below 400 hPa. The largest energy fluxes occur over the Greenland Sea (GS) (Sector 12), the Norwegian Sea (NS) (Sector 1) and the Chukchi Sea (Sector 7), where the air is warmer and contains more water vapour than over land. The landÁocean contrast is more pronounced for latent heat (Fig. 3a) than sensible heat (Fig. 3c). The sensible heat fluxes are greater than the latent heat fluxes by more than a factor of 2. The lowest fluxes are estimated for cross-sections (4 and 5) over Central Siberia and the mountainous region to the east of 1208E where the air is extremely cold and dry under persistent high pressures. More transient eddy activity occurs over the North Atlantic than other regions as indicated by the variances of v (Fig. 3g) and the mean positive local vorticities (Fig. 3i), contributing to the land-ocean contrast in eddy fluxes of energy.
Consistent decreases from 1979 to 2012 in transient eddy activities and associated energy fluxes are shown for crosssections 3 and 4 (608Á1208E) over regions south of the Kara Sea and Taymyr Peninsula in western and central Siberia ( Figs. 1 and 3). Winter blocking frequency was found to have increased for the same time period over western and central Siberia as shown by the time-series of average frequency for 408Á808N and 608Á1208E (Barnes et al., 2014). Less significant but apparent increases of northward energy fluxes are shown in Fig. 3b, d and f for cross-sections 10Á12 (2708Á 3608E, Fig. 1 Sorteberg and Walsh, 2008). Each section is referred to by a number (1Á12) and its adjacent ocean or land region. The dashed line indicates the Arctic Circle (66.68N). activity appears to have become stronger over the NS and the GS (sections 1 and 12, Fig. 3h and j).

Interannual variability of transport in the North Atlantic and Siberia
In this section, we examine the interannual variability as well as the long-term trend in transient eddies in the northern Atlantic and Siberia. Figure 4 shows the winter mean positive vorticity and the time fraction of high vorticity (TFHV: vort !1.0 )10 (5 s (1 ) in the mid-troposphere (approximately 400Á700 hPa) over the GS and the NS (308W to 308E) and over west-central Siberian (WCS) (608E to 1208E), respectively. Over the GS and NS, the mean vort is generally above the long-term average during 2000Á2012 and is lower during the earlier period (Fig. 4a)  shows large year-to-year variations but no clear trend (Fig. 4c). Over the WCS, mean vort is often above the long-term average in the early years, close to average in the middle years and below average in the later years (Fig. 4b).
These parameters show similar interannual variations and long-term trends with large correlation coefficients between each pair, except for the TOA LW flux. The skin temperature and the downward surface LW flux are also similar to those derived from the ERA-Interim reanalysis data for 1990Á2010 (Woods et al., 2013), whose interannual variations were found to be associated with intense moisture intrusions to the Arctic. Table 3 shows correlation coefficients calculated for mean positive vorticity with the Arctic mean climate parameters and for TFHV with the same parameters. Significant negative correlations (at p B0.01) are found between TFHV over the WCS and the downward surface LW flux (r 0(0.73 for TFHV 400Á700 hPa), precipitable water ( (0.69) and skin temperature ((0.67) (compare curves in Fig. 4d and 5). The correlation between TFHV and TOA LW flux is insignificant ( (0. 17). The correlation coefficients are smaller for 700Á1000 hPa TFHV and for mean positive vorticity in both layers. Similar correlations are insignificant over the northern Atlantic. These correlations suggest that the cyclonic circulation (or a lack of) over the WCS may vary with and even possibly plays a role in driving winter climate in the Arctic. The causes of circulation changes over Siberia are beyond the scope of this paper.

Summary
In this study we calculate the ANET to the Arctic based on meteorological data from the NCEP/NCAR reanalysis for the period of 1979Á2012. During this period the ANET decreased in all seasons, while the Arctic temperature increased and sea ice decreased. The downward trends are statistically significant for winter and annual mean fluxes, but not for other seasons. We calculated the energy fluxes, the mean variance of v and the mean positive local vorticity by cross-sections (400 hPa Á surface, every 308 from 08 to 3608E) for each winter. Linear regressions of the time-series show a general increase of transient eddy activity and associated energy fluxes over the northern Atlantic, and a general decrease over the west and central Siberia. Interwinter variations in transient eddy activity, as measured by the time fraction of high local vorticity, over the WCS are significantly anti-correlated with the Arctic climate.

Acknowledgements
Catherine Raphael drew Figure 1. Vaishali Naik, Michael Winton and Rong Zhang commented on the manuscript. Two anonymous reviewers provided constructive comments that led to significant improvements of the manuscript.