Characteristics of the 925 hPa Arctic air temperature field are evaluated with respect to controls by horizontal temperature advection, vertical motion, and characteristics of the underlying land and ocean surface. The highest winter mean temperatures in the Arctic, found in the Norwegian and Barents seas, are maintained primarily by cold horizontal advection countered by diabatic heating, the latter linked to open ocean waters. By comparison, temperatures over the central Arctic Ocean are primarily maintained by warm advection and diabatic cooling. Depending on the region, vertical motion opposes (notably off the east coast of Greenland) or reinforces forcing by advection. For summer, cold advection over snow-free land is countered by diabatic warming. This contrasts sharply with the Arctic Ocean, where warm advection and (locally) downward vertical motion combine to oppose pronounced diabatic cooling, the latter linked to surface melt and heat uptake in the ocean mixed layer. Prominent temperature anomalies in all seasons accompany onshore and offshore flow. For example, summer northerlies, blowing off the Arctic Ocean, yield cold anomalies over northern Eurasia extending far inland from the coast. While onshore winter westerlies yield above-average temperatures over northwestern Eurasia and the Barents and Kara seas, easterlies yield cold anomalies in the same regions. The most recent decade (2000–2009) has seen positive temperature anomalies over most of the Arctic for northerlies, easterlies, southerlies, and westerlies and for all seasons. Influences of recent shifts in atmospheric circulation, reduced sea ice extent, and rising sea surface temperature are prominent, especially for winter and autumn.
 The local rate of change of air temperature is controlled by horizontal temperature advection, vertical motion (the combination of vertical advection and adiabatic heating) and diabatic heating (radiative and turbulent heat exchanges). In the middle and higher latitudes, temperature change through horizontal advection is in large part associated with migrating extratropical cyclones and anticyclones that, aggregated in a zonal mean view, contribute to a net poleward energy transport. Characteristics of the surface also play a role, especially on regional and local scales. For example, while onshore (offshore) flow in summer tends to advect relatively cool (warm) air, the opposite relationship may hold during winter when the land is colder than the ocean. Vertical motion typically, but by no means always, opposes temperature change linked to advection. While the latitudinal gradient in solar heating of the surface ultimately drives the net poleward energy transport, diabatic heating is also linked to characteristics of the underlying surface, and to longwave radiative fluxes influenced by clouds and water vapor, sensible heat flux divergence and latent heat release. Relationships between winds, surface conditions and temperature can be expected to evolve in a nonstationary climate, through alterations in atmospheric energy transport requirements and surface conditions.
 The present paper examines how horizontal temperature advection, vertical motion, diabatic heating and surface conditions collectively shape spatiotemporal variations in the lower tropospheric (925 hPa) and surface air temperature fields of the Arctic. Use is made of atmospheric reanalysis data for the period 1979–2009 in conjunction with information on sea ice concentration, sea surface temperature and snow cover. An analysis of controls on seasonal mean temperature and its variability is followed by a focus on the most recent decade 2000–2009, which has seen outsized warming of the Arctic surface and lower troposphere relative to lower latitudes. While this Arctic amplification has been linked in large part to downward trends in sea ice extent and concentration fostering enhanced heat transfers from the Arctic Ocean to the atmosphere [e.g., Serreze et al., 2009; Screen and Simmonds, 2010], additional processes must be considered. These include how the wind field spreads out the effects of local warming from ice loss to surrounding regions [Serreze et al., 2009], changes in atmospheric circulation [e.g., Graversen et al., 2008], and the possibility that the Arctic atmospheric circulation is itself modified by the strong warming [Francis et al., 2009; Overland and Wang, 2010]. Our analysis helps to shed light on some of these issues.
 We employ 6-hourly wind and temperature data from the National Centers for Environmental Prediction/National Center for Atmospheric Research (NCEP/NCAR) reanalysis [Kalnay et al., 1996] which has been used in many recent studies of Arctic climate [e.g., Deser and Teng, 2008; Serreze et al., 2009; Overland and Wang, 2010]. Sea ice concentration anomalies are calculated from the satellite passive microwave record that combines time series from the Scanning Multichannel Microwave Radiometer and the Special Sensor Microwave/Imager using the National Aeronautics and Space Administration Team Algorithm [Fetterer et al., 2002]. Sea surface temperature (SST) anomalies are based on the National Oceanic and Atmospheric Administration (NOAA) Optimal Interpolation (OI) record V2, which relies on in situ and satellite SSTs, plus SSTs simulated by sea ice cover [Reynolds et al., 2002]. Snow cover anomalies are calculated from the Rutgers University Northern Hemisphere weekly snow extent product (http://climate.rutgers.edu/snowcover/index.php) [Robinson et al., 1993; Robinson and Frei, 2000].
 The period 1979–2009 corresponds to the passive microwave sea ice record and for which data from modern satellite data streams are available for ingest within the NCEP/NCAR reanalysis system. The NOAA OI SST record begins in late 1981. Primary use is made of temperature and wind analyses at the 925 hPa, based on assimilating observations (largely from satellite retrievals and rawinsondes) within output from the numerical weather prediction model at the core of the NCEP/NCAR system. We emphasize the analyzed 925 hPa temperature over surface air (2 m) temperature as the latter is strongly determined by the modeled surface energy budget; surface temperatures are used as appropriate to help assess the vertical structure of recent temperature anomalies. The NCEP/NCAR system uses observed sea ice from the passive microwave record, but only two states, ice covered and ice free, are considered, based on a 55% concentration threshold. A constant ice thickness of 2 m is assumed. Serreze et al.  discuss impacts of the sea ice treatment in NCEP/NCAR in comparison to the newer Japanese 25 year reanalysis, JRA-25 [Onogi et al., 2007]. A key conclusion is that the vertical, spatial and seasonal temperatures anomaly structures between NCEP/NCAR and JRA-25 are similar, with monthly anomalies in NCEP/NCAR (as assessed with respect to a 1979–2007 baseline) tending to be of larger magnitude. The NCEP/NCAR data are provided on a 2.5 × 2.5 degree latitude/longitude grid.
2. Climatological Relationships
Figure 1 shows the mean (1979–2009) Arctic temperature fields for winter (December–February) and summer (June–August) at the 925 hPa level and the surface. Regions where the 925 hPa level is below the local surface are masked; this masking is adopted in Figures 2–14. Hereafter, the Arctic region is defined as the region poleward of 60°N.
 The dominant features for winter are the warm conditions over the Atlantic side of the Arctic, contrasting with the extreme cold over northeastern Eurasia associated with the Siberian High, and over the northern Canadian Arctic Archipelago. At both the surface and at 925 hPa, the temperature field over the central Arctic Ocean is rather flat. The temperature distribution for summer is much more zonal in comparison to winter. Due to the melting ice surface, summer surface air temperatures over the Arctic Ocean are near the freezing point. Note the sharp summer temperature gradients along the coast, linked to differential heating of the atmosphere between the snow-free land and cold and mostly ice-covered Arctic Ocean [Serreze et al., 2001]. The fields for spring and autumn (not shown) depict the transition back to summer and winter, respectively.
 To assess how the mean seasonal 925 hPa temperature fields are maintained, use is made of the diagnostic equation for the local rate of change of temperature T:
The first term on the right is the horizontal temperature advection. The second term is the combined effect of adiabatic heating and the vertical temperature advection (hereafter referred to as the temperature change by vertical motion), where p is pressure, κ = R/cp = 0.286, R is the gas constant, cp is the specific heat of the atmosphere at constant pressure, and ω = dp/dt is the vertical velocity. In an adiabatic atmosphere (the vertical temperature gradient equals the dry adiabatic lapse rate, i.e., the vertical gradient in potential temperature is zero) the temperature change from vertical motion is also zero. The final term on the right is the diabatic heating rate.
 We computed the tendency, advection and vertical motion terms for each grid point at the 925 hPa level every 6 h. The tendencies at a given 6-hourly analysis time were computed as a centered difference using temperatures for the following and preceding 6 h analysis. Advection was calculated from u and v winds, and the horizontal gradient of air temperature. The temperature gradient was calculated using spherical harmonics. Omega is available as a calculated field in the NCEP/NCAR archives. The vertical temperature gradient was calculated from a centered difference using temperatures at the 850 hPa and 1000 hPa levels. Diabatic heating was computed for each 6 h time step as a residual from the three measured terms in equation (1). We then computed seasonal means of the terms.
 Some structure in the resulting fields will be lost as a result of the coarse resolution of the reanalysis. For example, one might expect to see more structure in the combined adiabatic and vertical advection terms (Figures 2 (bottom left) and 3 (bottom left)) over Alaska as a result of major topographic features. However, in other regions these combined vertical terms show expected features. For example, the expected large warming around the margin of Greenland that results from predominantly downslope low level flow off the ice sheet can be seen in Figures 2 and 3. Finite time and space differences used to calculate advection and temperature tendency can introduce errors. Because the diabatic term was calculated as a residual, errors accumulated in this field.
 The mean tendency for spring and autumn is on the order of about 0.2K per day, and is smaller for summer and winter. The mean tendencies in all seasons are in general more than an order of magnitude smaller than the other three terms. Hence the seasonal mean temperature can be viewed as approximately controlled by advection, vertical motion and diabatic heating.
Figure 2 provides winter fields of the temperature tendency, and contributions to temperature change by horizontal advection, vertical motion, and diabatic heating. Units are K per day. Several features stand out. The Norwegian and Barents seas, which have the highest mean winter temperatures in the Arctic at the 925 hPa level (Figure 1), are dominated by negative (cold) horizontal advection. This indicates that this region tends to lie on the western side of cyclones moving poleward along the North Atlantic storm track (see winter map of cyclone center counts from Serreze and Barrett [2008, Figure 1]). Cold advection is in turn largely countered by diabatic warming. This warming is consistent with the presence of open ocean water which in winter extends as far as 80°N in this sector. In lying within the North Atlantic storm track, this region is furthermore characterized by extensive cloud cover which helps to limit radiative heat loss to space. Warm advection also dominates much of Eurasia, which tends to be countered by diabatic cooling, linked to the cold land surface and more limited cloud cover. Cold advection is also prominent along the east coast of Greenland, and while reinforced by diabatic cooling, it is strongly countered by warming associated with downward vertical motion. By contrast, most of the central Arctic Ocean is characterized by warming through horizontal advection, generally countered by diabatic cooling. The effects of vertical motion are more localized.
 Corresponding fields for summer follow in Figure 3. Positive temperature advection contributes to warming over the northern North Atlantic and the Arctic Ocean, while negative advection contributes to cooling over the Eurasian landmass and northwest North America. In general, the temperature tendency associated with horizontal advection is countered by the contribution from the diabatic term. In explanation, while cold advection over land contributes to cooling at the 925 hPa level, the snow-free surface readily absorbs solar radiation so that the overlying air is heated by convection, and longwave radiation. Warming through convection is furthermore favored in that cold advection tends to reduce static stability. There is a summer precipitation maximum over the Arctic lands (and much of this is convective precipitation), implying a significant contribution to diabatic heating from latent heat release. By comparison, while horizontal advection contributes to warming over the ocean, solar energy is used largely to melt sea ice and warm the ocean mixed layer, meaning less heating of the overlying atmosphere. These contrasts are evident in the sharp discontinuity along the coast in the mean July net surface heat flux shown in the Arctic energy budget study of Serreze et al. . Over coastal land, the net surface heat flux is typically 10–20 W m−2 (into the ground) compared to 100 W m−2 along the coastal seas. This contrast is also manifested in the sharp coastal temperature gradient seen in Figure 1. This sharp gradient corresponds to the summer Arctic frontal zone remarked upon in a number of past studies [e.g., Reed and Kunkel, 1960; Serreze et al., 2001]. That one does not see this coastal temperature gradient for winter follows in that the ice-covered Arctic Ocean and land are both cold. An alternative explanation is that diabatic heating over land in summer sets up a temperature gradient between land and ocean, which drives temperature advection. Both explanations are plausible.
 Diabatic cooling in summer is most pronounced in a band along the coastal seas. This feature is colocated with a band of warming through vertical motion, which implies descent in a statically stable environment (potential temperature increasing with height). While positive temperature advection is found over all of the Arctic Ocean, there are also local peaks along the coastal seas adding to the warming from vertical motion. The pronounced diabatic cooling is consistent with these processes; while warming through advection and downward motion favor more radiative cooling at the 925 hPa level, these dual processes also favor strong ice melt and heat uptake in the ocean mixed layer. Positive temperature advection by itself will lead to upward motion of air parcels along isentropic surfaces. A reasonable explanation is the effects of differential vertical motion in a frontal zone that act to oppose the horizontal advection. As outlined by Holton , this vertical-plane circulation is characterized by ascent on the warm side of the front and descent on the cold side of the front (the ocean side of the Arctic frontal zone), acting to weaken the temperature gradient (tilting the isentropes from vertical to horizontal).
 Fields of temperature advection, vertical motion and diabatic heating for the spring and autumn seasons (not shown) have features of winter and summer. A feature of all seasons is the dominance of warm advection over the Arctic Ocean, countered by diabatic cooling. In spring and autumn, as in winter, cold advection over the Norwegian Sea is countered by diabatic warming. Spring also shares with summer the cold advection over Eurasian and North American land areas that is countered by diabatic warming. However, in autumn, the pattern of warm advection over Eurasia countered by diabatic cooling has more in common with winter. Cold advection over North America, countered by diabatic warming is a feature in all seasons.
 We next ask the question: what is the temperature anomaly structure over the Arctic with respect to winds from different directions? To this end, using 6-hourly temperatures and wind components at 925 hPa, seasonal mean temperature anomalies at each grid point over the period 1979–2009 were computed separately for winds with components from the south, north, east and west. For example, the temperature anomaly for southerly (from the south) winds for the period 1979–2009 is defined as:
where T′s79–09 is the 6-hourly grid point temperature for cases for which the wind has a southerly component (the meridional wind is positive), θs79–09 is the angle of the wind vector for winds with a southerly component measured clockwise from due easterly (from the east), T79–09 is the temperature irrespective of wind direction and n79–09 is the number of 6-hourly observations irrespective of wind direction. The adjustment by sinθs gives full weight to winds blowing directly from the south, and a smaller weight to winds closer to due easterly or westerly, and can be thought of as the “degree of southerliness” of the winds. For example, an easterly wind (no southerly component) is assigned the angle 0° and a sine weighting of 0, and a westerly wind (also no southerly component) is assigned an angle 180° and also a sine weighting 0. By comparison, a southerly wind is assigned an angle 90° and a sine weighting of 1. Similar results are obtained by simply defining southerly winds using all observations for which winds were blowing from between 135° and 215°. Expressions similar to equation (2) were used to calculate temperature anomalies for winds blowing from the north, east and west.
 The plots for winter (Figure 4) reveal the expected pattern of opposing temperature anomalies for northerly and southerly winds that together contribute to a net poleward transport of atmospheric energy. Opposing temperature anomalies for northerly and southerly winds are particularly strong (locally −5 and +5 K) over the Atlantic side of the Arctic. This is a reflection of eddy heat transports along the northern end of the north Atlantic cyclone track. There is a clear correspondence between the area of strong negative temperature anomalies associated with northerly winds and the area of strong positive temperature anomalies associated with southerly winds in the Norwegian Sea. This area of contrasting anomalies is located just south of the highly baroclinic climatological margin of the wintertime sea ice cover. This area also corresponds to the peak in cold advection shown in Figure 2. This peak in cold advection is not inconsistent with the results shown in Figure 4 and simply indicates that northerly winds, and their associated cold anomalies, occur more often in this region than the other wind directions.
 Westerly winds in winter are associated with positive temperature anomalies over most of northern Eurasia and its coastal seas from near the prime meridian to 120°E, strongest in the Kara and northern Barents seas (+3 K). This anomaly structure points to the influence of air masses that have been warmed and moistened by passage over the open waters of the northern North Atlantic. Negative temperature anomalies for the same region when the winds are easterly reflect advection from winds blowing from the cold continent and Arctic Ocean. Similarly, maritime (continental) easterlies (westerlies) yield positive (negative) temperature anomalies over eastern Eurasia. Note also how marine easterlies are associated with strong positive anomalies over the Atlantic side of Greenland and in Davis Strait. The pattern for spring (not shown) is similar to that for winter, but with generally weaker temperature anomalies, pointing to seasonal weakening of the primary storm tracks and smaller temperature contrasts between the land and ocean.
 The situation for summer (Figure 5) is quite different. The strong opposing temperature anomalies for northerlies and southerlies associated with the north Atlantic and Pacific cyclone tracks dominating the winter pattern are no longer present. The dominant feature instead is opposing zonally oriented temperature anomalies along the Eurasian coast and extending eastward along the coast of Alaska and western Canada. This manifests the contrast between the warm, snow-free land and the cold Arctic Ocean discussed earlier. When winds are from the north, they advect cold air over the Arctic Ocean toward the land, yielding strong negative temperature anomalies along the coast and smaller negative anomalies that extend far inland. Winds from the south by contrast originate from the much warmer snow-free land yielding positive temperature anomalies along the Arctic coast. Summer temperature anomalies for westerly and easterly winds are in general smaller in summer than in winter, particularly over the Arctic Ocean where the melting ice surface subdues regional temperature contrasts. However, maritime easterlies (continental westerlies), yield negative (positive) anomalies over eastern Eurasia. This is the expected result of the land tending to be warmer than the ocean; in winter, the anomaly pattern is reversed. For much of the rest of northern Eurasia, summer easterlies are linked to positive temperature anomalies. Broad bands of contrasting anomalies for easterly and westerly winds occur along the coasts of Eurasia and Alaska in summer which may be related to the passage of cyclones in that easterly winds precede warm frontal passage and are warmer than westerly winds that follow cold frontal passage.
 Temperature anomaly patterns by wind direction for autumn (not shown) document the transition back toward winter conditions, with reestablishment of opposing anomalies over the Atlantic and Pacific sides of the Arctic associated with the primary storm tracks, and contrasting anomalies for easterlies and southerlies linked to warm maritime (compared to land) versus cold continental (compared to the ocean) source regions.
3. Recent Temperature Anomalies
 Building from the introduction, Arctic amplification which has emerged over the past decade is an expected response of our planet to positive radiative forcing [e.g., Manabe and Stouffer, 1980; Holland and Bitz, 2003]. Model-projected Arctic amplification is often discussed in the context of albedo feedback, a process which appears to have been first articulated by Croll  and Arrhenius . Over land, the view is fairly straightforward. Warming leads to earlier loss of the spring snow cover, which then leads to solar heating of the darker underlying surface, expressed as a higher air temperature. However, projected Arctic amplification in simulations from most climate models of recent vintage is strongest not over land, but over the Arctic Ocean. While a number of processes contribute, including advection and feedbacks linked to water vapor and cloud cover that influence the downwelling longwave radiation [Alexeev et al., 2005; Solomon, 2006; Winton, 2006; Graversen and Wang, 2009; Deser et al., 2010], a key driver is reduced sea ice extent at the end of the summer melt season, in part a seasonally lagged expression of albedo feedback.
 Briefly, as the climate warms, the sea ice melt season lengthens and intensifies, leading to less sea ice at summer's end. Summertime absorption of solar energy in dark open water areas increases the sensible heat content of the ocean mixed layer, contributing further to ice melt. Ice formation the following autumn and winter, important for insulating the warm ocean from the cooling atmosphere, is delayed. This promotes strong upward heat fluxes, seen as strong warming at the surface and in the lower troposphere. Projected Arctic amplification is hence less prominent in summer, when energy is used to melt sea ice and increase the sensible heat content of the ocean mixed layer [Serreze and Francis, 2006].
 Since the beginning of the modern satellite record in October 1978, the extent of Arctic sea ice has declined in all months, but with the strongest downward trend (11% per decade through 2009) at the end of the melt season in September. A record low was set in September 2007 [Stroeve et al., 2008; http://nsidc.com/arcticseaicenews/]. The ice cover is also thinning [Nghiem et al., 2006; Maslanik et al., 2007]. Using data from the NCEP/NCAR and JRA-25 reanalyses, Serreze et al.  demonstrate consistency between the seasonal and spatial structure of recent Arctic air temperature anomalies and the effects of sea ice loss; the key element of the vertical structure pointing to an anomalous surface heat source is that the warming is strongest at the surface. This seasonality and vertical structure is also well captured in vertical cross section plots of temperature trends for 1989–2008 based on data from ERA Interim [Screen and Simmonds, 2010], the newest reanalysis from the European Centre for Medium-Range Weather Forecasts (http://www.ecmwf.int/products/data/archive/descriptions/ei/index.html).
 We can get some sense of the impacts of atmospheric circulation on the pattern of recent temperature anomalies by looking at the advection of the climatological mean (1979–2009) temperature by the anomalous wind for 2000–2009:
where 〈T〉 is the mean 925 hPa temperature averaged over all 6-hourly analyses for the period 1979–2009 and V′ is the wind anomaly, expressed as the average over all 6-houly analyses for 2000–2009 minus the mean for the period 1979–2009. A similar calculation was made by Deser and Teng  for the period 1979–2007. Effects of anomalous winds are most evident in winter (Figure 7). The prominent positive temperature anomaly between Svalbard and Novaya Zemlya (Figure 6) clearly corresponds to a region of warm advection by the anomalous wind (approximately 1 K per day). More specifically, the 925 hPa mean wind for the 2000–2009 period has an anomalous southerly wind component in this area, blowing across the mean isotherms computed for 1979–2009 (see Figure 1). In turn, the one area with negative anomalies in winter temperature, from about 120°E to the date line, can be linked to winds with an anomalous northerly component, advecting cold air into the region.
 However, from comparing Figures 6 and 7, it is clear that advection by the anomalous wind provides but a partial explanation of the temperature anomaly pattern for winter. The same can be said of other seasons. For example, while the negative temperature anomalies for 2000–2009 in spring that dominate the quadrant from about 180–90°W can be related to cold advection by the anomalous wind, anomalous cold advection also dominates the quadrant from the prime meridian to about 90°E, where the temperature anomalies are positive.
 It is instructive to examine the vertical structure of recent temperature anomalies. If the temperature anomalies at 925 hPa are primarily driven by an anomalous diabatic heating from the surface (e.g., open water, reduced surface albedo), one expects that the air temperature anomalies at the surface will be larger than those at the 925 hPa level [Serreze et al., 2009; Screen and Simmonds, 2010]. As a simple way of expressing the vertical structure, we computed for each season the ratio of surface to 925 hPa temperature anomalies for the most recent decade 2000–2009 (Figure 8). If the ratio is larger (smaller) than one, the surface temperature anomaly is larger (smaller) than the anomaly at 925 hPa. Areas where either the surface or the 925 hPa anomaly is less than 0.25 K are masked in Figure 8. While there are large areas in all seasons with ratios larger than one, pointing to surface forcing, there are also large areas with ratios less than one, pointing more to effects of atmospheric circulation and other factors. These results must be viewed with the caveat that that the 925 hPa temperature is an analyzed field whereas surface temperature is calculated from the surface energy budget.
 To visualize sources of anomalous surface forcing, Figure 9 provides seasonal anomaly maps of sea ice concentration, SST, and weekly snow cover frequency for the 2000–2009 decade. NOAA SSTs are plotted for open ocean regions, defined here as outside of areas where ice concentration is at least 15% for at least 50% of the time over the period 2000–2009. Negative anomalies in ice concentration over the past decade for winter and spring are focused over the Atlantic sector of the Arctic, compared to the more widespread negative anomalies for summer and especially autumn. Positive SST anomalies characterize almost all open ocean regions north of 60°N. In general, anomalies in the frequency of snow cover (the % of weeks in a season with snow cover) are modest. While the pattern over Eurasia is dominated by reduced snow cover frequencies in spring, frequency changes exceeding 5% lie primarily south of 60°N. High-latitude land areas in central Eurasia and eastern Siberia, and also much of Baffin Island have negative anomalies in the frequency of snow cover in summer. These reductions in summer snow cover frequency occur in early June, reflecting an earlier snowmelt. July and August have little or no snow cover in these regions. For both North America and Eurasia poleward of 60°N, snow cover frequency anomalies for the 2000–2009 decade are primarily positive in autumn, but changes for individual grid cells are generally less than 7.5%. However, different snow cover data sets yield different results. Brown et al.  examined trends in Arctic (north of 60°N) snow cover extent (coverage in square km) during the spring melt period (May–June) from ten different data sources over the period 1967–2008, and found a more linear reduction in spring extent than shown in the NOAA charts. From a multidata set approach, they computed a 46% reduction in Arctic snow covered area in June and a 14% reduction in May over the period 1967–2008.
 Anomalies of diabatic heating at 925 hPa for winter for the 2000 to 2009 decade provide further insight to sources of anomalous warming (Figure 10). The diabatic heating term for the 2000 to 2009 period was calculated following the approach describe for the long-term means shown in Figure 2. Anomalies for the 2000 to 2009 period were calculated with respect to long-term mean of diabatic heating shown in Figures 2 and 3. The patterns of anomalies present a complex picture which is difficult to interpret. The largest anomalies occur in winter. Two strong positive anomalies stand out in the diabatic heating field in this season between Svalbard and Novaya Zemlya and over the Arctic Ocean between 150° E and 90° W, as well as over the Canadian Arctic Archipelago. The positive anomaly between Svalbard and Novaya Zemlya is the result of a more positive diabatic heating in the most recent decade. The positive anomaly on the opposite side of the Arctic is the result of less negative diabatic cooling in this region.
 While as just discussed, the prominent positive temperature anomaly in winter between Svalbard and Novaya Zemlya (Figure 6) can be linked to anomalous horizontal advection (Figure 7), the 925 hPa temperature anomaly is also underlain by a larger surface anomaly (the ratio of the surface to 925 hPa locally exceeds 1.8) (Figure 8) and a strong positive diabatic heating anomaly (Figure 10), that are roughly colocated with negative sea ice anomalies (Figure 9). While it can be concluded that both changes in atmospheric circulation and surface conditions contribute to the 925 hPa temperature anomaly, the southerly wind anomaly also likely plays a role in maintaining the anomalous open water in the first place both through the wind stress on the ice (wind stress acts to push the ice edge poleward of its usual position) and positive temperature advection (discouraging winter ice growth). It may also involve ocean heat flux convergence [Chylek et al., 2009]. Similarly, the strong winter temperature anomalies in Baffin Bay and the eastern Canadian Arctic Archipelago appear to have contributions from both anomalous advection (Figure 7) and reduced ice concentration (Figure 9).
 Turning to spring, 925 hPa temperature anomalies in the region between Svalbard and Novaya Zemlya, while positive, do not show up as a local maximum. However, that the temperature anomaly is stronger at the surface points to a continued role of surface forcing linked to negative sea ice anomalies. The diabatic heating anomaly in this region in spring is positive. Temperature advection by the anomalous mean wind is actually negative in this region (Figure 7). Stronger spring temperature anomalies are found over Baffin Bay and Davis Strait, but the forcing appears to be complex with local contributions from both the surface, linked to negative sea ice concentration anomalies and anomalous advection. The positive 925 hPa temperature anomalies in the sector from 90°E to the date line, with a local maximum over the East Siberian Sea, seem to be more closely allied with a broad area of warm advection by the anomalous mean wind; sea ice concentration anomalies are small, and there are no pronounced anomalies in the frequency of occurrence of snow cover (Figure 9).
 The most prominent feature of the 925 hPa temperature anomaly field for summer (Figure 6) is a band of modest (>1 K) positive anomalies along the Eurasian coast from the Laptev to the East Siberian Sea. While this band of positive 925 hPa temperature anomalies is underlain by negative anomalies in sea ice concentration, the ice is melting and solar energy is being used (in open water areas) to raise the sensible heat content of the ocean mixed layer. As such, one expects a much more limited surface influence on the 925 hPa temperature fields than would accompany the same concentration anomalies for winter. The diabatic heating anomalies are smaller in this season. While the pattern of 925 hPa temperature anomalies is certainly suggestive of a surface influence, the surface to 925 hPa ratio in temperature anomalies is less than one. Note also that locally, the anomalous horizontal advection both contributes to the positive 925 hPa temperature anomaly (in the East Siberian Sea) and opposes the anomaly (in the Laptev Sea).
 Autumn 925 hPa temperature anomalies for 2000–2009 are positive across the entire domain north of 60°N, strongest over the Arctic Ocean. While the ratios of the surface to 925 hPa temperature anomaly point to a strong surface forcing over the Chukchi and East Siberian seas, where negative anomalies in autumn sea ice concentration in recent years have been particularly pronounced, ratios exceeding unity also encompass most of the Arctic Ocean as well as parts of the land area. A strong positive diabatic heating anomaly is colocated with the positive temperature anomaly centered about 170°E. There is also a local maximum in the ratio over the Canadian Arctic Archipelago. Note that the positive temperature anomalies over land occur in conjunction with positive anomalies in the frequency of occurrence of snow cover. While we cannot provide a clear explanation for this result, we speculate that it may reflect greater cold season precipitation occurring as a result of a warmer atmosphere.
4. Changing Wind and Temperature Relationships
 To examine how relationships between anomalies in temperature and winds summarized in Figures 4 and 5 have changed over the past decade, mean monthly and seasonal 925 hPa temperatures for each grid point stratified by each of the four cardinal wind directions for the period 1979–2009 were subtracted from weighted seasonal means for the most recent decade, 2000–2009. Taking the example of southerly winds as was done in equation (2), we have:
where T′s00–09 is the mean temperature anomaly for 2000–2009. The second term in the right hand side is the same as in equation (2), i.e., the mean temperature for southerly wind events for 1979–2009, with the individual 6-hourly observations weighted by the “degree of southerliness” represented by sinθ s. The first term on the right side, the mean temperature for southerlies for the period 2000–2009, contains an additional weight ϕsinθ, which is the ratio of the percent frequency of observations with a wind direction of sinθ for 2000–2009 with respect to 1979–2009. This ratio was computed for 10 bins of sinθ; if sinθ for a given 6-hourly observation fell within the bounds of a given bin, it was assigned the ratio for that bin. The intent of the weighting is to adjust for differences in atmospheric circulation (“degree of southerliness”) between the 2000–2009 and 1979–2009 periods. Similar equations were used to calculate temperature anomalies for the other cardinal wind directions. From our calculations, we can answer the question: Have temperatures associated with southerly, northerly, easterly or westerly winds changed compared the 1979–2009 climatology?
 Results aggregated by month for the region poleward of 60°N and an embedded Arctic Ocean domain appear in Figure 11. The Arctic Ocean domain is the same as that used in the energy budget study of Serreze et al. ; it excludes the Canadian Arctic Archipelago, Baffin Bay/Davis Strait, the Norwegian Sea and the East Greenland Sea. The regional mean anomalies are positive for every month and for each wind direction, most strongly expressed in October through December. Anomalies are in general larger for the embedded Arctic Ocean domain.
 Spatial fields of temperature anomalies for winter follow in Figure 12. Similar patterns are obtained when anomalies are calculated without the ϕsinθ weighting from equation (4). The Arctic region is dominated by positive temperature anomalies for all wind directions, indicating a large-scale background warming that is consistent with greenhouse gas forcing. Nevertheless, there is considerable spatial structure. Recall that winter features a prominent warm anomaly for the 2000–2009 decade at both 925 hPa and the surface centered between Svalbard and Novaya Zemlya (Figure 6). Providing further evidence that this regional feature is strongly driven by anomalous surface heating, namely, a negative anomaly in sea ice extent, the temperature anomaly appears for all wind directions. For example, while background warming makes northerlies, southerlies, easterlies and westerlies warmer than they used to be, winds from all directions blowing over anomalous open water are also warmed from the surface upward. The peak temperature anomaly for southerlies is less prominent than the anomaly linked to northerlies. This makes sense, for while cold northerlies are moving over a relatively much warmer surface, southerlies are warmer to begin with (again see Figure 4), will tend to maintain their warmth in their poleward journey by positive SST anomalies, and will hence tend to be warmed less by the anomalous surface heat source associated with reduce ice concentration. Temperature anomalies between Svalbard and Novaya Zemlya are in turn stronger for westerlies as compared to easterlies. Why this is the case is not entirely clear. One might argue that air parcels moving over the anomalous open water from the west originate over the cold ice-covered ocean, and are hence warmed more strongly by the anomalous surface heat source than air parcels moving in from the east.
 Nevertheless, as winds from all directions appear to be warmed by the anomalous surface heat source, it follows that the effects of the surface heat source will also tend to be spread out horizontally (downwind of whatever direction the wind is blowing). This helps to explain why the peak temperature anomaly between Svalbard and Novaya Zemlya (Figure 6) is embedded within a much broader region of strong positive temperature anomalies encompassing much of the Atlantic side of the Arctic. Similar arguments for winds from all directions being warmed in response to reductions in ice concentration and higher SST explain the relatively strong winter warming encompassing Baffin Bay/Davis Strait. By sharp contrast, winds from all directions, particularly southerlies, are locally colder in the most recent decade over eastern Eurasia. This is the only region in winter with negative winter temperature anomalies over the past decade (Figure 6), which as discussed can be related to an anomalous cold advection (Figure 7).
 Results for spring (not shown) are also characterized by the predominance of positive anomalies for all wind directions in comparison with the 1979–2009 climatology. However, consistent with the anomaly pattern in Figure 2, winds from all directions are locally colder during the most recent decade from the date line eastward to about 90°W. Summer temperature anomalies are again primarily positive for all wind directions (Figure 13), with some evidence that northerlies in particular are being warmed by open water along the Eurasian coastal seas. All winds for the most recent decade are cooler over parts of the area from the date line eastward to 90°W. It is difficult to cleanly interpret the anomaly structure for autumn (Figure 14); what we appear to be seeing is a complex pattern of background warming, atmospheric circulation spreading out the effects of anomalous surface heating, and other factors.
5. Summary and Discussion
 The climatological mean (1979–2009) Arctic temperature field at the 925 hPa level is maintained by interplay of horizontal advection, vertical motion and diabatic heating. The highest temperatures in the Arctic, over the Norwegian and Barents seas, are maintained primarily by cold horizontal advection countered by diabatic heating, the latter linked to open ocean waters. Temperatures over the colder central Arctic Ocean are primarily maintained by warm advection and diabatic cooling. Depending on the region, vertical motion variously opposes or reinforces the forcing by advection. For summer, cold advection over the snow-free land is countered by diabatic warming, while over ocean, warm advection and (locally) downward vertical motion combine to oppose pronounced diabatic cooling, the latter linked to surface melt and heating of the ocean mixed layer. Turning to anomalies, summer northerlies, blowing off the Arctic Ocean, yield colder than normal conditions over northern Eurasia; this advection extends far inland from the coast. While onshore winter westerlies yield above-average temperatures over northwestern Eurasia and the Barents and Kara seas, easterlies yield cold anomalies in the same regions.
 Arctic temperature anomalies for the most recent decade 2000–2009 are interpreted here as reflecting the combined effects of (1) a general background warming which is part of the planet's response to positive radiative forcing; (2) anomalies in atmospheric circulation; and (3) changes in characteristics of the surface, in particular, reduced sea ice extent and higher SSTs compared to climatology. While our study has not addressed factors such as warming by black carbon aerosols, soot on snow and changes in cloud cover atmospheric and water vapor, we acknowledge that they may be quite important. A major challenge is providing information on the spatial structure of these other factors that could allow for meaningful comparisons with fields of advection, diabatic heating, and vertical motion.
 Background radiative forcing is suggested from the widespread warming for the recent decade that is present for all seasons and for temperature anomalies stratified by each of the four cardinal wind directions. Anomalies in atmospheric circulation introduce spatial structure to the seasonal temperature anomaly patterns. For example, strong positive temperature anomalies centered between Svalbard and Novaya Zemlya in winter owe their existence in part to an anomalous southerly wind component at 925 hPa level. Circulation (in particular anomalous northerlies) also explains local cooling, such as seen during spring over the quadrant from the date line eastward to 90°W. The effects of reduced ice concentration are most apparent as regional “hot spots” in the 925 hPa temperature anomaly field. Large ratios (exceeding one) of the surface to 925 hPa anomalies indicate greater warming at the surface expected with a surface heat source.
 Processes can be mutually supporting. The best example is the positive temperature anomaly between Svalbard and Novaya Zemlya in winter: while both wind stress and warmth associated with anomalous southerly winds help to maintain open water, vertical heat fluxes from the open water help to keep the atmosphere warm. An observed change in ocean circulation characterized by enhanced transport of warm Atlantic waters into the Arctic Ocean [Polyakov et al., 2005], discouraging winter ice growth along the Atlantic ice margin, may also be involved.
 With regard to the trajectory of the Arctic system through the 21st century, an important issue is how the effects of atmospheric warming due to reduced sea ice extent and concentration and higher SSTs will be spread out by winds to affect surrounding regions, acting as a feedback to foster more ice melt and reduce ice growth, or leading to enhanced warming over land affecting the vegetation and soil temperature regime [e.g., Lawrence et al., 2008]. For the period 2000–2009, effects of winds in “spreading out the heat” is apparent over the Atlantic side of the Arctic in winter. However, in other seasons, the evidence so far is less clear.
 This study was supported by National Science Foundation grants ARC-0901962 and ARC-0805821.