Notice: Wiley Online Library will be unavailable on Saturday 30th July 2016 from 08:00-11:00 BST / 03:00-06:00 EST / 15:00-18:00 SGT for essential maintenance. Apologies for the inconvenience.
 A current paradox is that many physical and biological indicators of Arctic change—summer sea-ice extent, spring surface air temperature and cloud cover, and shifts in vegetation and other ecosystems—show nearly linear trends over the previous two and a half decades, while the Arctic Oscillation, a representative atmospheric circulation index often associated with Arctic change, has had a different, more episodic behavior, with a near-neutral or negative phase for 6 of the last 9 years (1996–2004) following a positive phase (1989–1995). Stratospheric temperature anomalies over the Arctic, which serve as an index of the strength of the polar vortex, also show this episodic character. Model projections of Arctic temperature for 2010–2029 show model-to-model and region-to-region differences suggesting large variability in the future response of atmospheric circulation to external forcing. Thus internal processes in the western Arctic may have a larger role in shaping the present persistence of Arctic change than has been previously recognized.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 Change in the Arctic has continued over the previous 25 years in many indicators: warmer surface air temperatures, reduced sea-ice area, conversion of tundra to dense shrubs and wetland, and biological consequences. Yet the Arctic Oscillation (AO), an indicator of Arctic circulation and often considered to be part of the explanation for such changes, has shown a more episodic character (multi-year runs of positive or negative anomaly values). The AO, an indicator of primarily winter sea-level pressure—but also keyed to stratospheric variability [Thompson et al., 2000], has been near-neutral or negative for 6 of the previous 9 years, after a period of persistent positive anomalies in the early 1990s. Thus there is an apparent paradox between the continuing climatic change trends in many surface-based indicators and the behavior of the AO. A further issue is that model projections of future Arctic change show impacts from reduced sea ice primarily in winter [Arctic Climate Impact Assessment (ACIA), 2004], while recent changes often occur in spring and summer.
2. The State of the Arctic Oscillation
 The time series of the winter (November–March) AO through 2004 is shown in Figure 1. Feldstein  noted that the trend of the 3-month version of the winter (DJF) AO from 1968 to 1997 was 0.057 yr−1, which was greater than the 99% confidence level expected for 30-year trends, based on synthetic time series developed from the 1958–1997 AO record. He concludes that the trend for this period is beyond that expected from internal variability of the atmosphere and is “due to either coupling to the hydrosphere/cryosphere or to driving external to the climate system.” We repeated the Feldstein calculation for the more recent 30-year period of 1975–2004 and found a trend of 0.018 yr−1, which is near the center of the distribution of expected AO trends developed by Feldstein [2002, Figure 5]. By inspection of Figure 1, the low AO values in 1969 and 1970 and the high values in the 1990s suggest that the 1968–1997 period noted by Feldstein indeed may have been part of the population of AO 30-year trends but near the extreme value, as its magnitude was not sustained by later data.
 Stratospheric temperature anomalies at 200 hPa are related to the strength of the polar vortex and the AO [Pawson and Naujokat, 1999]. The cold anomalies in the early 1990s in March signaled a continuation of a strong polar vortex into spring (Figure 2). In contrast, the period of the 1980s had warm springtime anomalies and the period after 1998 was more variable. The magnitude of cold stratospheric temperatures in spring over the Arctic, while not an exact match of the AO, has its episodic character.
3. Recent Arctic Change
 The northern hemisphere areal extent of sea ice in September (Figure 3, left) shows a nearly linear trend of ice area decrease of almost 20% over the previous two and a half decades, primarily in the western Arctic. Recent ice extents represent a 300–500 km retreat from median values (Figure 3, right). Of particular interest is that the previous 3 years (2002, 2003, 2004) show year-to-year persistent low values, in contrast to previous low values which could be considered as 1-year extremes such as (1995).
 Turning to surface air temperature anomalies at Arctic weather stations relative to a 1961–1990 base period, Tromso and Tiksi (central Siberia) had warm temperature anomalies and Egedesminde (west Greenland) was cold in winter beginning in 1989, in accord with temperature advection anomalies associated with the positive AO (Figure 4). There are some warm anomalies in North America around 1980 and Siberia in mid-1980s, but these appear to be localized in region and duration. Of particular interest are the warm anomalies in west Greenland and Tromso after 1996, and the recent warming throughout northern North America, which do not appear to map onto the AO associated patterns of temperature anomalies. In spring there are isolated temperature anomalies throughout the record, but the major warm period is near Bering Strait eastward to continental North America beginning in the late 1980s. The warm temperature anomalies in spring at Barrow in the early 1990s can be shown to be related to the new presence of storms (compared to the 1980s) as a result of the breakdown of the AO-polar vortex [Overland et al., 2002]. There are nearly Arctic-wide positive anomalies in spring for 2002–2004, which are especially strong in the Beringia region (Shimidta–Barrow). Maps of surface temperature trends for 1982–2003 (Figure 5) during winter (DJF) and spring (MAM) based on AVHRR data [after Comiso, 2003] show similar spatial patterns, highlighting NE Canada in winter and the Chukchi Sea in spring.
 Other Arctic indicators tend to show linear trend behavior since the late 1970s (www.arctic.noaa.gov/detect). Satellite estimates of greenness (NDVI) can be calibrated against Arctic land type base maps to indicate approximate regions of tundra [Wang and Overland, 2005]. These calculations show an estimated 17% loss of tundra area from the late 1970s to the late 1990s. The north Atlantic and Bering Sea are showing a general replacement of Arctic species by sub-Arctic species [Beaugrand et al., 2002; Overland and Stabeno, 2004]. Greenland continues to show strong shrimp harvests (United Nations, http://faostat.fao.org).
 We are left with an apparent paradox of more linear Arctic climate change beginning in the late 1970s and the more episodic AO. It has been argued that the spatial patterns of the atmospheric circulation response to external forcing may in fact project principally onto modes of natural climate variability such as the AO [Corti et al., 1999]. The lack of a continued upward trend in the AO over the last decade, while not ruling out this hypothesis, also suggests a weak red noise response [Wunsch, 1999]. Portmanteau and cumulative periodogram tests for white noise based on the 1951–2004 AO (JFM) reject the hypothesis at the 90% level, thus supporting a red-noise stochastic nature for the “episodic” behavior in the AO.
 It is instructive to investigate the surface air temperature (SAT) and sea level pressure (SLP) fields in the recent years of near-neutral AO (Figure 6). Three representative years are shown. What is apparent is the large spatial extent of positive SAT anomalies in recent years compared to the years of near-neutral and negative AO values in the 1960s and 1980s. The SLP anomaly patterns in March are most striking in their amplitudes with ridge and trough (high and low SLP) patterns over the Arctic, which are responsible for temperature advection contributions to regional SAT anomalies in April. Yet a key feature of both the recent SAT and SLP fields is their major year-to-year variability in geographic distribution of the anomalies. Thus part of the answer to the paradox is that strong pressure gradients over the Arctic in the last decade are maintaining SAT anomalies through advection.
 The western Arctic may be having a larger role in shaping the persistence of Arctic change than has been previously recognized; note the region of sea-ice anomalies in the western Arctic in Figure 3 (right) and the location of warm spring temperature anomalies in Figures 4 and 5. Rigor and Wallace  show that areal coverage of thick multi-year ice decreased precipitously during 1989–90 when the AO was in an extreme high index state. Under these conditions anomalies of thin ice can recirculate within the western Arctic providing a multi-year signal. With longer ice-free periods during summer to absorb heat energy, we may be in a state where first-year sea ice cannot grow to its historical thickness due to a positive feedback [Lindsay and Zhang, 2005]. Other examples are shifts in vegetation [Sturm et al., 2005] and the increase in spring (March–May) cloud cover over the Arctic seas [Schweiger, 2004]. A second answer to the paradox is that the nearly 20% changes in surface conditions (sea ice and land) are sufficient to help maintain persistence of current trends.
 A conceptual model (Figure 7) may help the discussion of intrinsic versus forced variability in Arctic change. It shows external forcing in the subtropics, a mid-latitude polar vortex and Arctic feedbacks. Although the polar vortex has two positive feedbacks, one through dynamic cooling of the stratosphere and one through modification of lower tropospheric winds, the interaction between the time-mean flow and transient eddies in the atmosphere [De Weaver and Nigam, 2000] argues for a large climate noise paradigm, i.e., internal variability, affecting the Arctic. Some models demonstrate an Arctic response to external forcing involving subtropical-Arctic connections through the stratosphere [Johannessen et al., 2004; Rauthe et al., 2004], although there is still concern about understanding mechanisms (D. Thompson, personal communication). Direct evidence for Arctic/subtropical connections comes from a positive phase AO type temperature response in the winters following volcanic eruptions in the subtropics [Robock and Mao, 1992]. Thus external forcing can conceivably change the distribution of Arctic circulation anomalies such as the AO, but with large regional and temporal variability. While model projections for 2040–2059 show the eventual impact of warmer temperatures in winter from major reductions in sea ice, projections for 2010–2029 show much more variability [ACIA, 2004]. Figure 8 shows the change in February surface temperature (years 2010–2029 minus 1980–1999) for two of five models presented in the ACIA Report; note the strong spatial variability of warm and cold anomalies and their different locations in different models. Regional and temporal variability should not be unexpected for the next several decades given external and internal forcing of Arctic circulation.
 The variability between models and their ensemble members supports the concept that the year-to-year and decade-to-decade variability in the AO does not rule out the steady influence of external forcing. Without the persistent positive phase of the AO and given the relatively short records, however, clear attribution of forced versus intrinsic Arctic change based solely on observed atmospheric circulation is difficult. On the other hand, the current trends in many other Arctic indicators are disturbing, and will require human adaptation to these changes over the coming decades. The experience of the last decade suggests considerable humility in addressing Arctic change.
 We appreciate the support of the NOAA Arctic Research Office. We thank J. Comiso for allowing us to present Figure 5. D. Percival performed the white noise test on the AO data series. We appreciate discussions with M. Serreze, J. Walsh, D. Shindell, D. Rind, and D. Thompson. PMEL Contribution 2761, JISAO Contribution 1111.