3.3.1. EOF1

Heterogeneous correlation maps between the expansion coefficient Series of EOF1 and climatological field anomalies (SLP, Sst, Sat, winds) are presented in figure 9. on these maps, large positive (negative) correlation indices denote Strong positive (negative) correlations between EOF1 and normalized anomalies of the climatological fields (e.g. SLP). the correlation pattern for SLP anomalies (fig. 9a) Shows a Strong opposition between the greenland Sea and a latitudinal band Stretching along the 40˚n parallel from newfoundland to western europe. Strongest correlation values occur over newfoundland (0.32) and the denmark Strait (–0.39), which is the normal position of the icelandic low. this dipole pattern of SLP anomalies resembles the positive phase of the arctic oscillation (AO; Reference Thompson and WallaceThompson and wallace, 1998), and even more closely the north atlantic oscillation (NAO; Reference HurrellHurrell, 1995). this was confirmed by correlating the winter (december–february) portions of the EOF1 and nao index Series, which gave r = 0.54 at lag 0. a weaker but Significant negative correlation was also found between EOF1 and the el niño–southern oscillation (ENSO) monthly SLP index at lag +3 months (i.e. ENSO leading by 3 months). these results are consistent with observations by Reference Wang, Mysak and IngramWang and others (1994) and Reference Mysak, Ingram, Wang and van der BaarenMysak and others (1996), who found positive Sea-ice anomalies in hb and the labrador Sea when Strong episodes of ENSO occurred in the previous Summer and during positive phases of the nao during winter of the Same year.

Fig. 9.

Heterogeneous correlation maps (lag 0) between EOF1 and monthly anomalies of (a) SLP; (b) u- and v-wind components; (c) SAT; and (d) SST. Color bars and arrow length represent the Strength of the correlation coefficient. Black contours delineate areas of Statistical Significance (p < 0.05). indices (fig. 13a) revealed a Statistically Significant relationship with pna at lag 0 (r = 0.16), with ENSO at lag +4 months (r = 0.22) and with the nino3 index at lag +5 months (r = 0.34). the better correlation with the nino3 and ENSO indices indicates that the link between EOF3 and the pna is mostly due to ENSO fluctuations. the variability of EOF3 over time is dominated by a few high-amplitude events, which are often preceded by an ENSO event (fig. 13b). the cross-correlation function with ENSO and nino3 Suggests an oscillation with a period of ∽2–4 years, consistent with the timescale of ENSO variations (2–6 years) and with the low-frequency part of EOF3.

Fig. 10.

Same as Figure 9 but for EOF2.

Fig. 11.

Same as Figure 9 but for EOF3.

Fig. 12.

Same as Figure 9 but for EOF4.

Fig. 13.

(a) Cross-correlation plot between EOF3 expansion coefficients and monthly indices values of the PNA, ENSO and Nino3 indices. Note: El Niño events are defined by a negative ENSO index and positive Nino3 index, while the PNA tends to be inversely correlated with ENSO. (b) Plot of EOF3 expansion coefficients (black curve) and Nino3 index (grey curve). Black dots represent NAO events below one Standard deviation of the monthly NAO index (only the maximum monthly value within the extreme year is Shown).

Changes in the Strength and distribution of SLP anomalies may affect Sea ice through dynamic (wind) and thermodynamic (air-temperature) forcing. the correlation map of EOF1 with wind anomalies (fig. 9b) Shows that an anomalous cyclonic circulation over greenland combined with anticyclonic circulation over newfoundland induces Strong northwesterly wind anomalies in the outermost part of the DS/LS region (fig. 9b), favouring a Southward advection of ice in winter and Spring. the northwesterly wind anomalies also cause advection of cold arctic air into the eastern canadian arctic, hb and labrador Sea, promoting the growth and preservation of positive Sea-ice anomalies in these areas (fig. 9c). the correlation pattern for Sat (fig. 9c) is consistent with the Spatial Signature of a positive nao, with cold and dry winters in northern canada and greenland (Reference HurrellHurrell, 1995). EOF1 also Shows Strong negative correlations with Sst over the north atlantic region (fig. 9d). lag correlation maps between EOF1 and Sst (not Shown) Show negative Sst anomalies appearing in Southwestern greenland 1 year before the positive ct anomaly develops in the DS/LS and hb regions. the Sst anomaly progressively expands eastward into the norwegian Sea up to 6 months after the ct anomaly, and Southward toward the newfoundland coast where it dissipates 1–1.5 years after the ct anomaly. the lag-correlation maps between EOF1 and Sat also Show Strong negative Sat anomalies developing 1 year before EOF1 but decaying a few months later (not Shown). this is in good agreement with recent results Showing that decadal cold climatic episodes and positive ice-cover anomalies around greenland and baffin are accompanied by cold Sst anomalies lasting 1–3 years after the decay of ice-cover anomalies (Reference Rogers, Wang and McHughRogers and others, 1998; Reference Deser, Holland, Reverdin and TimlinDeser and others, 2002).

When normalizing the correlation coefficients by their estimated large-lag Standard errors, Sat (σ _max = 5.2σ) appears most Strongly related to EOF1, followed by SLP and winds (σ _max ∽ 4.6σ) and Sst (σ _max = 3.8σ). these results Suggest that atmospheric temperatures contribute more to the ct anomalies expressed by EOF1 than dynamic forcing. these conclusions complement recent results by Reference Prinsenberg, Peterson, Narayannan and UmohPrinsenberg and others (1997) and Reference Deser, Holland, Reverdin and TimlinDeser and others (2002) who found that positive anomalies develop in the DS/LS region in response to anomalous northwesterly winds within the labrador Sea and offshore flow east of newfoundland. a Subsequent model Simulation by Reference Deser, Holland, Reverdin and TimlinDeser and others (2002) Suggested, as reported here, a dominance of thermodynamic processes over dynamic mechanisms.

3.3.2. EOF2

Significant but weaker correlations occur between EOF2 and SLP anomalies, with maximum negative polarity over the labrador Sea (r = –0.21) and maximum positive polarity (r = 0.17) over western and Southern north america (fig. 10a). lag-correlation maps (not Shown) Suggest that negative SLP anomalies lead positive EOF2 anomalies by 1–2 months in the DS/LS region, while the Smaller positive SLP anomaly appears farther South 2 months earlier. the associated wind pattern Shows northerly wind anomalies over much of the eastern canadian arctic, and Southeasterly and easterly wind anomalies over davis Strait (fig. 10b). correlation with Sat Shows maximum values over the DS/LS region (r = 0.42) and minimum values (r = –0.25) over eastern canada/USA (fig. 10c). positive Sat anomalies lead the positive EOF2 anomalies by 3–4 months, gradually expanding over the DS/LS region and dissipating 1 month after the ct anomaly, while negative Sat anomalies only appear at lag 0 and dissipate 1 month after the ct anomaly (not Shown). closer inspection of the expansion coefficient Series of both EOF1 and EOF2 (fig. 7) reveals that although both Series are Statistically uncorrelated as a whole, EOF2 includes much of the annual to Sub-annual variability of EOF1. their winter Structure is Similar, being inversely correlated at r = –0.99. the winter portion of EOF2 is in turn also correlated with the winter NAO, but negatively (r = –0.55). So while EOF1 represents a predominately decadal Signal most active during the cold Season, EOF2 appears to be an annual to Sub-annual expression of this first mode, which is related to the winter NAO. it is thought that the dipole between the DS/LS region and HB, which is Superimposed on EOF1, results from local ice-to-atmosphere feedbacks controlled by the coupled interannual variability of the NAO and ice cover in the DS/LS region. reduced ice cover in the DS/LS region during a more negative NAO allows for turbulent heat transfer from the ocean to the atmosphere, causing local warming and decreased SLP (fig. 10a and c). the fact that the Sat anomalies associated with EOF2 appear 1–2 months before the SLP anomaly Supports this idea. the local SLP anomaly induces northerly winds and advection of cold air over part of HB, promoting positive ice anomalies in that region. the reverse process occurs when ice cover increases during winter in the DS/LS region in response to a more positive NAO. this hypothesis is Supported by recent results Showing high upward turbulent heat-flux anomalies occurring in the DS/LS region when negative ct anomalies occur in winter (Reference Deser, Walsh and TimlinDeser and others, 2000, Reference Deser, Holland, Reverdin and Timlin2002). when taking the Standard errors of correlation into account, the onshore Surface winds and positive temperature anomalies in the DS/LS are the most Significant feature associated with EOF2 (r >4σ). the Sst pattern of figure 10d, which is only present at lag 0, may represent a direct effect of ct variability associated with EOF2: decreased (increased) Ssts occur in areas of increased (decreased) Sea-ice cover. alternatively, the correlation may reflect the effect of the parameterization between Sst and Sea-ice concentration in the ERSST database (Reference Smith and ReynoldsSmith and reynolds, 2004).

3.3.3. EOF3

Slp anomalies associated with EOF3 Show a weak dipole with positive polarity over the arctic ocean (r = 0.20) and negative polarity over the bering Sea (r = –0.15) (fig. 11a). the positive SLP anomaly developed 3–4 months earlier over the arctic ocean, while the negative anomaly over the bering Sea appeared 1 month earlier (not Shown). the Surface wind pattern associated with EOF3 Shows anomalous offshore winds along the beaufort Sea coast and anomalous northerly and Southerly winds over nares Strait and baffin island, respectively (fig. 11b). correlation with Sat Shows a warming over northwestern north america and cooling over the north pacific and northwestern USA (fig. 11c). the Sst pattern Shows warming in the beaufort Sea (fig. 11d), which began 3–4 months before the ct anomaly and disappeared 2 months later (not Shown). when normalizing the correlation by their Standard error, Sst appears as the dominant factor (r = 6σ), followed by wind and Sat (r = 3.6σ), and SLP (r = 3.3σ).

The positive ctanomaly in baffin bay Seems to be a direct result of the anomalous anticyclonic circulation over the arctic ocean, which promotes advection of ice into baffin bay through nares Strait (fig. 11b). at the Same time, the anomalous Southerly winds over baffin island may restrain the ice from moving South, thereby maintaining the positive ct anomaly. the most prominent feature of EOF3, however, is the negative ct anomaly off the beaufort Sea coast (see fig. 6c). the anticyclonic SLP anomaly over the arctic appears to be causing most of the dynamical forcing on Sea ice, by inducing offshore wind anomalies which promote export of ice from the beaufort Sea toward the eastern arctic (fig. 11b). on the other hand, the cyclonic anomaly over the bering Sea acts as a thermodynamic forcing on Sea ice by driving warm pacific air into the beaufort Sea and western coast of north america (fig. 12b and c). the position of the two centres of SLP variability is typical of late-spring to early-summer conditions, when the beaufort high and the aleutian low begin to dissipate. because the Sstanomalies are locally restrained to the beaufort Sea region, they are more likely to act as a positive feedback on Sat and ice cover rather than being the main driver of the ct anomalies. it is likely that preconditioning of the ice pack during Spring by enhanced offshore wind and positive temperature anomalies triggered the decrease in ct, with positive Ssts developing Simultaneously and further enhancing the decrease in ct through thermodynamic effects.

The SLP and Sat patterns of figure 11 have Similarities with the pacific/north american teleconnection pattern (pna; Reference Barnston and LivezeyBarnston and livezey, 1987). the PNA is one of the most prominent modes of low-frequency variability in the northern hemisphere. its positive phase is associated with above-average SLP in the vicinity of hawaii and over the inter-mountain region of north america, and below-average SLP over the aleutian islands and over the Southeastern united States. increased (decreased) Sat occurs over northwestern north america and the north pacific, respectively. although the pna is a natural internal mode of climate variability, it is Strongly affected by ENSO, with positive phases of the pna being associated with ENSO years (Reference Stocker and HoughtonStocker and others, 2001, p. 453). cross-correlation between EOF3 and the pna, ENSO (SLP) and nino3 (sst)

The positive SLP anomaly over the arctic differs from the pna pattern and Suggests an influence from the north atlantic (fig. 11a). a weak but Statistically Significant negative correlation was found between EOF3 and the monthly nao index at lag +1 month (r = –0.16). however, the largest anomalies in EOF3 occur when nao and ENSO events occur Simultaneously (fig. 13b). the most Striking example is the extreme year of 1998 during which the largest ENSO event over the Study period occurred in combination with the Second largest nao event. the amplitude of the ct anomaly appears to depend on the Synchronicity of both indices. a weakening of the beaufort gyre occurs during winter/spring under a negative nao regime, which tends to increase ice export from the beaufort Sea (Reference KwokKwok, 2000). thus large negative ct anomalies in the beaufort Sea result from preconditioning of the ice pack during winter–spring combined with the thermodynamic and dynamic forcing induced by ENSO events. these results complement recent findings by Reference Maslanik, Serreze and AgnewMaslanik and others (1999) who found Similar SLP anomaly patterns from composite charts associated with light ice years in the beaufort Sea. they Suggested that both the nao and ENSO could have an impact on ice anomalies in the area but found no clear relationships. the present application of the EOF method allowed for the extraction of the variance in the beaufort Sea region associated with these two modes of climate variability.

3.3.4. ;EOF4

The SLP pattern associated with EOF4 (fig. 12a) Shows a weak positive anomaly over central canada (r = 0.16) resulting in anomalous northwesterly winds over hb, while anomalous Southerly winds occur along the eastern coast of baffin island in response to a weak decrease in SLP over Southern baffin island (fig. 12b). Similar Sat and Sst patterns are found, with increased temperature over the western coast of north america and bering Sea and reduced temperature over the central north pacific (fig. 12c and d). decreased Sat also occurs over Southern hb (fig. 12c). the Strongest correlations between EOF4 and Sst occur within hb (r = 0.46), but Since these only appear at lag 0, they most likely reflect the effect of the parameterization between Sst and Sea-ice concentration in the ersst database (Reference Smith and ReynoldsSmith and reynolds, 2004).

The Spatial patterns of SLP, Sat and Sst are Strongly apparent to the east pacific–north pacific teleconnection index (ep-np; Reference Barnston and LivezeyBarnston and livezey, 1987). the ep-np is a Spring–summer–fall pattern whose positive phase is associated with a Southward Shift and intensification of the pacific jet Stream from eastern asia to the eastern north pacific, followed downstream by enhanced anticyclonic circulation over western north america and by enhanced cyclonic circulation over the eastern united States. the positive phase of the EP-NP pattern is associated with above-average Surface temperatures over the eastern north pacific, and below-average temperatures over the central north pacific and eastern north america. maximum correlation between EOF4 and the monthly EP-NP (r = 0.21) occurs when the EP-NP leads EOF4 by 1 month. on a Seasonal basis, Significant correlations were found between EOF4 and the EP-NP during Spring (r = 0.63) and Summer (r = 0.57), in accordance with the Seasonality of both EOF4 and the EP-NP. when Scaling all correlations by their Standard errors and ignoring the correlation between EOF4 and Sst in hb, Sst and Sat appear to have a Stronger influence on EOF4 (r=3σ) than does SLP and winds (r=2.5σ). EOF4 thus appears to result from Springtime climate variability in the north pacific region, which induces anomalous anticyclonic circulation over central/ western north america. the resulting Southward advection of cold arctic air over hb retards ice melt in the Spring while the northwesterly winds tend to push the ct anomalies toward the Southern portion of hb. the weaker dipole in the DS/LS region may result from the opposition between the Southerly winds anomalies along the eastern coast of baffin island and the northwesterly wind anomalies over labrador.