1. Introduction
Climate regime shifts, often characterized by transitions between climate states, have been observed and investigated in many studies (e.g., Namias 1969; Nitta and Yamada 1989; Trenberth 1990; Graham 1994; Minobe 1997; Mantua et al. 1997; Watanabe and Nitta 1999; Deser et al. 2004; Lo and Hsu 2008; Swanson et al. 2009; Hansen et al. 2010). For example, global mean temperature has been found to decease during the 1900s and 1910s, increase during the 1920s and 1930s, decrease again from the 1940s to 1976/77, and increase from that point to the beginning of the twenty-first century (e.g., Swanson and Tsonis 2009; Hansen et al. 2010). The climate regime shifts have been related to the large-scale decadal–interdecadal climate variability over the Pacific and Atlantic Oceans (e.g., Trenberth and Hurrell 1994; Mantua et al. 1997; Overland et al. 1999; Watanabe and Nitta 1999; Deser et al. 2004; Swanson et al. 2009; Lo and Hsu 2010). The decadal–interdecadal climate variability, on the other hand, appears to be driven mainly by stochastic forcing, with ocean–atmosphere feedback playing a relatively minor role (e.g., Hasselmann 1976; Mantua and Hare 2002; Xie and Carton 2004; Latif et al. 2006; Power and Colman 2006; Liu and Alexander 2007; for a comprehensive review, see Liu 2012). In particular, decadal–interdecadal climate variability in the Pacific is associated with changes in wind-driven upper-ocean circulation, while multidecadal variability in the North Atlantic is mainly associated with changes in the Atlantic meridional overturning circulation (Liu 2012).
On continental and subcontinental scales, a climate regime shift around 1990 has been documented and investigated. Examples of this climate regime shift include the abrupt warming in northern East Asia in association with the snow cover decrease in Eurasia in the late 1980s (Watanabe and Nitta 1999), a regime shift in the Pacific decadal oscillation around 1989 (Hare and Mantua 2000), a reversal of the Arctic Oscillation in the late 1980s (Rodionov and Overland 2005) accompanied by abrupt warming in the extratropical Northern Hemisphere (Lo and Hsu 2010), a significant trend increase in the tropical western Pacific sea level in the early 1990s relative to the preceding 40 years (Merrifield 2011), and wintertime large-scale cooling trends across the eastern United States and northern Eurasia over the last two decades (Cohen et al. 2012) compared to mostly opposite temperature trends over 1961–90 (Yu and Lin 2012). In general, the exchange of midlatitude and polar air supports the anomalous temperature. The Eurasian snow cover during autumn has been found to play a role as an amplifier in the wintertime climate change around 1990 (e.g., Watanabe and Nitta 1999; Cohen et al. 2007; Solomon et al. 2007; Fletcher et al. 2007; Allen and Zender 2011; Peings et al. 2013). In addition, numerical experiments indicate that extratropical atmospheric forcing dominates over tropical forcing for the temperature and circulation trends over 1991–2010, while both extratropical and tropical forcings contribute to the trends over 1961–90 (Yu and Lin 2012). Nevertheless, it remains to explore whether the climate regime shift around 1990 is part of decadal–interdecadal climate variation or just a singular event.
This study focuses on several issues related to the northern wintertime temperature and circulation variations on decadal–interdecadal time scales. In particular, we analyze the decadal covariability of extratropical temperatures over the northern lands and atmospheric circulations over the Northern Hemisphere using two reanalysis datasets and a long preindustrial control simulation from the Canadian Earth System Model. We then relate the change of temperature and circulation trends around 1990 to the leading modes of the decadal covariability and examine the influence of synoptic eddies on the upper-tropospheric circulation and lower-tropospheric temperature anomalies. In addition, we analyze the relationship between the decadal temperature and circulation variability with several well-known atmospheric variability regimes.
The rest of the paper is organized as follows. The data and methodology are described in section 2. The annual and decadal temperature and geopotential variations are analyzed and compared in section 3. The decadal temperature and circulation covariability is described in section 4, based on two reanalysis datasets for the period 1951–2010, together with its relationship to the change of temperature and circulation trends around 1990 and the association with eddy–mean flow interaction and sea surface temperature anomalies. The decadal covariability result is further assessed in section 5, using the reanalysis data over the twentieth century and a long climate model simulation. The relationship between the temperature and circulation variability and several well-known atmospheric regimes is also presented in section 6. A summary is provided in section 7.
2. Data and methodology
a. Data
The analysis is mainly based on the December–February (DJF) monthly atmospheric temperature and circulation data extracted from both the National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR) reanalysis (hereafter NCEP-1) (Kistler et al. 2001) and the Twentieth-Century Reanalysis (hereafter 20C) (Compo et al. 2011) for the period 1951–2010, with relatively more reliable fields than longer reanalysis time series. Years are labeled according to the January dates in this analysis. Different from NCEP-1, 20C was generated by assimilating only surface pressures and using monthly SST and sea ice distributions as boundary conditions within an ensemble Kalman filter. We use the 56-member ensemble-mean monthly temperature and circulation data for 20C. The daily temperature at 850 hPa and horizontal winds at 250 hPa from NCEP-1 are used to examine the synoptic-scale eddy statistics. In addition, the monthly SST from 1951 to 2010 was obtained from the Met Office Hadley Centre Sea Ice and SST dataset, version 1.1 (HadISST1.1) (Rayner et al. 2003). All variables are interpolated to a standard 2.5° × 2.5° grid.
The monthly North Atlantic Oscillation (NAO) (see Hurrell et al. 2003, and references therein) index and Southern Oscillation index (SOI) (e.g., Peixoto and Oort 1992) from 1951 to 2010 were obtained from the National Oceanic and Atmospheric Administration/Climate Prediction Center (NOAA/CPC) (http://www.cpc.ncep.noaa.gov/data/indices). The monthly North Pacific index (NPI) (Trenberth and Hurrell 1994) for the same period was retrieved from the NCAR Climate Analysis Section (CAS) (http://climatedataguide.ucar.edu/guidance/north-pacific-indexnpi-trenberth-and-hurrell monthly-and-winter). The monthly Pacific decadal oscillation (PDO) (Mantua et al. 1997) and Atlantic multidecadal oscillation (AMO) (e.g., Schlesinger and Ramankutty 1994; van Oldenborgh et al. 2009) indices were obtained from the Joint Institute for the Study of the Atmosphere and Ocean (JISAO) (http://jisao.washington.edu/pdo/PDO.latest) and the NOAA/Earth System Research Laboratory (ESRL) (http://www.esrl.noaa.gov/psd/data/timeseries/AMO), respectively.
The 20C ensemble mean data for the long period from 1901 to 2010 and a 996-yr preindustrial control run output from the second-generation Canadian Earth System Model (CanESM2) (http://www.ec.gc.ca/ccmac-cccma/default.asp?lang=En&n=3701CEFE-1; Arora et al. 2011) are also used to examine the stationarity of the decadal covariability results obtained from the period 1951–2010 and to explore the long-term behavior of the decadal covariability between temperature and circulation anomalies. The atmosphere in the CanESM2 (McFarlane et al. 2005; von Salzen et al. 2013) comprises a 63 wave triangularly truncated (T63) spherical harmonic expansion with 35 layers from the surface to the stratopause region (1 hPa). The ocean model was developed from the NCAR Climate System Model (CSM) Ocean Model (NCOM) (e.g., Gent et al. 1998), using a z-level vertical coordinate with horizontal differencing formulated on an Arakawa B grid. There are 40 vertical levels with spacings ranging from 10 m near the surface to nearly 400 m in the deep ocean. Horizontal coordinates are spherical with grid spacings approximately 0.94° latitude and 1.41° longitude. A full description of the CanESM2, including coupling, vertical mixing, and other aspects, can be found at the above website.
b. Diagnostic methods
To assess the temperature and circulation variability, the following procedure is applied to the data. For the time series from 1951 to 2010, the linear trend over the 60 DJFs is subtracted from the original data at each grid point to remove the secular variation. This is also intended to partially remove the anthropogenic influence on climate (e.g., the greenhouse-gas-induced global warming) from the analysis. The linear trend for the time series of interest is calculated using normalized orthogonal polynomial approximations (e.g., Hildebrand 1956, chapter 7). The Kolmogorov–Zurbenko (KZ) filter is then employed to assess the decadal–interdecadal variability (Eskridge et al. 1997; Overland et al. 1999). The KZ filter is based on an iterative moving average that removes high-frequency variations from the time series of interest. We apply a KZ(7, 2) filter, in which a time series is smoothed with a 7-yr running mean applied twice, to filter out all periods less than 10 yr. The results reported here depend only weakly on a reasonable variation of the low-frequency filter [e.g., using KZ(6, 2) and KZ(8, 2) filters]. We apply the KZ(7, 2) filter to the 60 DJFs, together with a 3-DJF cutoff period at both ends of the output. Thus, the filtered time series contains 54 DJFs from 1954 to 2007 and is termed as the decadal DJF data.
The maximum covariance analysis (MCA) (Bretherton et al. 1992; Wallace et al. 1992) is employed to characterize the spatiotemporal pattern of maximum covariance between temperature and circulation anomalies. The relationship between a time series of interest and the associated atmospheric circulation and temperature anomalies is quantified through linear regression and multivariate linear regression. The power spectrum of time series is calculated using the Parzen estimator (e.g., von Storch and Zwiers 1999). We also examine the geopotential eddy activity associated with stationary wave anomalies in the upper troposphere and eddy heat fluxes associated with temperature anomalies in the lower troposphere. Analysis of the feedback by synoptic eddies on geopotential and temperature anomalies helps in the understanding of the maintenance of atmospheric circulation and temperature anomalies (e.g., Lau 1988; Trenberth and Hurrell 1994; Kug and Jin 2009; Choi et al. 2010). Synoptic-scale eddy statistics are obtained by applying a Butterworth bandpass filter to retain fluctuations within 2–8 days (Murakami 1979).
3. Temperature and circulation variability
Figure 1 displays the geographical distribution of the standard deviation of interannual DJF mean surface air temperature Ts, the standard deviation of the decadal filtered DJF mean Ts, and the ratio of the decadal standard deviation to the interannual standard deviation over the northern extratropics (20°–90°N). The DJF mean temperature variance exhibits a pronounced land–sea contrast with high variability over land, which is due to the different heat capacities of land and ocean. Variance of Ts is also small in low latitudes and large in mid- and high latitudes, with peak values over Canada, northern Eurasia, and the Siberian Arctic. The spatial distribution of the decadal Ts variance generally resembles that of the interannual variance. The ratios of the decadal Ts standard deviation to its interannual counterpart are 42.9% (39.1%), 42.5% (40.3%), and 43.6% (55.1%) for NCEP-1 (20C), averaged over the northern extratropical lands, ice-free oceans, and sea ice–covered areas, respectively. This indicates that the decadal surface temperature variance generally accounts for about 20% of the interannual variance in winter.
Standard deviations of (left) DJF mean surface air temperatures Ts and (center) decadal filtered DJF mean Ts and (right) the ratio of decadal filtered DJF standard deviation to DJF standard deviation over the northern extratropics (20°–90°N) for (top) NCEP-1 and (bottom) 20C. In (left)–(right), Contour intervals are 0.6°C, 0.4°C, and 10%, respectively. Values greater than 40% are hatched in the right panels.
Citation: Journal of Climate 27, 2; 10.1175/JCLI-D-13-00266.1
The temperature variance patterns are broadly similar in both reanalyses. Nevertheless, the 20C result is generally smoother than that of the NCEP-1, as would be expected owing to the 56-member ensemble mean used in the 20C. The main differences between the two reanalyses are seen over the Arctic sea-ice-covered area. Variances in Ts are larger over the Arctic regions spanning the East Siberian Sea to the Canadian Basin (120°E–90°W) and smaller over the region spanning from eastern Canada to the Siberian Arctic in 20C compared to NCEP-1, particularly for the decadal variance. The decadal Ts variance is also slightly smaller over northern Eurasia in the 20C than NCEP-1.
Figure 2 shows the corresponding variance of 500-hPa geopotential heights Φ500 over the northern extratropics. The DJF mean Φ500 variance is characterized by maxima over the north Pacific and Atlantic Oceans and over the Siberian Arctic, consistent with the result based on the earlier National Meteorological Center (NMC) reanalysis for the 18 DJFs from 1963 to 1980 (Wallace and Blackmon 1983). The locations of high geopotential variance also correspond closely to the regions of the midlatitude jet stream and storm tracks (e.g., Peixoto and Oort 1992). The decadal Φ500 variance pattern bears resemblance to its interannual counterpart. However, the ratio of the decadal Φ500 standard deviation to the interannual deviation varies over the northern extratropics, relatively low over the polar region and the North Pacific. The area averaged ratios for NCEP-1 (20C) are 44.6% (37.0%), 41.2% (35.0%), and 29.5% (28.6%) over the extratropical lands, ice-free oceans, and sea-ice-covered areas, respectively. In addition, the Φ500 variance pattern also has good correspondence in the two reanalyses, although the 20C result is slightly smoother than the NCEP-1 result.
As in Fig. 1, but for 500-hPa geopotential heights: contour interval 100 (50) m2 s−2 in the left (center) panel.
Citation: Journal of Climate 27, 2; 10.1175/JCLI-D-13-00266.1
4. Decadal temperature and circulation covariability
In this section, we examine the covariability of the decadal land surface temperature and atmospheric circulation anomalies over the northern extratropics and analyze its relation to the change of temperature and circulation trends around 1990, to the feedback of synoptic eddies on the upper-tropospheric circulation and lower-tropospheric temperature, and to the SST anomalies.
a. Decadal covariability
The maximum covariance analysis between the decadal land surface temperature and 500-hPa geopotential height anomalies over the northern extratropics (20°–90°N) was performed for both NCEP-1 and 20C over the 54 decadal-filtered DJFs. The two reanalyses exhibit similar results of the first two MCA patterns, which dominate the total squared covariance. Hence, Fig. 3 displays the two leading patterns of combined MCA of decadal Ts and Φ500 anomalies over the northern extratropics and the corresponding expansion coefficients using the decadal filtered fields from 1954 to 2007 in both NCEP-1 and 20C. Statistics related to the MCA analysis are listed in Table 1. These two leading MCA modes explain 89.5% of the total squared covariance and are well separated from the next most important covariance mode (Table 1). The two leading MCA modes explain comparable fractions of total decadal Ts variance (33.1% versus 31.6%, Table 1) and Φ500 variance (38.0% versus 25.6%) and are also the two principal modes of decadal land surface temperature variability over the northern extratropics as determined by an empirical orthogonal function analysis (not shown).
(top) First two modes of combined MCA between Ts (colors, °C) and Φ500 (contours 50 m2 s−2, negative values hatched) over the northern extratropics for the 54 decadal filtered DJFs from 1954 to 2007. (bottom) The corresponding MCA expansion coefficients of decadal Ts for NCEP-1 (black) and 20C (blue) and Φ500 for NCEP-1 (red) and 20C (magenta) anomalies.
Citation: Journal of Climate 27, 2; 10.1175/JCLI-D-13-00266.1
The squared covariance fraction (SCF; %) for the first two modes of the combined MCA expansion, the temporal correlation coefficient between the expansion coefficients of the decadal Ts and Φ500 anomalies over the northern extratropics, and the explained fraction (%) to the total variance by the leading modes for individual fields.
The positive phase of the leading MCA pattern (MCA1) (Fig. 3, top left) is characterized by positive temperature anomalies over most of Canada, the eastern United States, Mexico, and Eurasian mid- to high latitudes and by negative temperature anomalies over the western United States, northeast Canadian Arctic, Greenland, North Africa, and parts of the Eurasian subtropics. The 500-hPa geopotential height exhibits negative anomalies over the polar region, North Atlantic high latitudes, and the North Pacific, as well as positive anomalies elongated zonally from the northeast Pacific toward North America, North Atlantic midlatitudes, and then Europe and over Asian midlatitudes. The exchange of midlatitude and polar air following the anomalous circulation, which implies the temperature advection, supports the temperature anomalies. The MCA1 pattern also bears resemblance to the leading MCA pattern between wintertime seasonal mean sea level pressure and surface temperature (including SST) over the Northern Hemisphere from 1899 to 1993 (Hurrell 1995), which is closely related to the decadal circulation and temperature anomalies over the North Atlantic and surrounding landmasses.
The expansion coefficients of decadal Ts and Φ500 anomalies (Fig. 3, bottom left) for the MCA1 are broadly similar in the NCEP-1 and 20C datasets and reveal transition points in the early 1960s, the late 1970s, and the early 2000s. Nevertheless, the amplitude of the expansion coefficients of Ts and Φ500 anomalies is slightly weaker in 20C than in NCEP-1, indicating that the MCA1-associated temperature and geopotential variability is relatively weaker in 20C compared to NCEP-1. Again, this is probably due to the ensemble mean applied in the 20C data.
The positive phase of the second MCA pattern (MCA2) (Fig. 3, top right) is dominated by positive temperature anomalies over most of North America, Eurasia, and Greenland and by negative temperature anomalies over the west coasts of Europe and North Africa, the East Asian midlatitudes, and patches of cold anomalies over the southeast United States. The consistency of the temperature anomalies with the atmospheric circulation anomalies is also apparent. The 500-hPa geopotential height exhibits high anomalies over North Atlantic, North American mid- to high latitudes and central Eurasia, with low anomalies between these regions. The anomalous circulation may alter the temperature advection, leading to the temperature anomalies described above. The expansion coefficients of Ts and Φ500 anomalies (Fig. 3, bottom right) for the MCA2 reveal transition points in the late 1960s, 1970s, 1980s, and 1990s for both reanalyses, although the changes in the late 1970s and 1980s are weak for Ts in NCEP-1.
b. Temperature and circulation trends
The two leading MCA patterns (Fig. 3, top) of the decadal land surface temperature and 500-hPa geopotential anomalies bear some resemblance to the Ts and Φ500 trends over 1961–90 and 1991–2010, as discussed in Yu and Lin (2012) and Cohen et al. (2012). This resemblance is further supported by the MCA expansion coefficients of Ts and Φ500 anomalies (Fig. 3, bottom), which reveal a shift of temperature and circulation trends around 1990 on interdecadal time scale, about 20–30 yr. Figures 4 and 5 display the interdecadal-scale linear trends of Ts and Φ500 calculated using the decadal DJF data and the corresponding trends reconstructed using the first (first two) MCA mode(s) over 1961–90 (1991–2007). Table 2 lists the spatial correlation coefficients relating the trends calculated from the three different approaches within the domain of interest in Figs. 4 and 5.
(left) Linear trends of Ts (colors, °C decade−1) and Φ500 (contours 100 m2 s−2 decade−1, negative values hatched) from 1961 to 1990 over the northern extratropics and corresponding trends reconstructed using the (center) first MCA mode and (right) first two MCA modes: (top) NCEP-1 and (bottom) 20C.
Citation: Journal of Climate 27, 2; 10.1175/JCLI-D-13-00266.1
As in Fig. 4, but from 1991 to 2007.
Citation: Journal of Climate 27, 2; 10.1175/JCLI-D-13-00266.1
Spatial correlation coefficients between the linear trend calculated using the decadal filtered data (Trd0) and the corresponding trends reconstructed using the first MCA mode (Trd1) and the first two MCA modes (Trd2) for Ts and Φ500 over the northern extratropics (20°–90°N) in the two periods.
The temperature and geopotential height trends over the two periods (left panels in Figs. 4, 5) bear close resemblance to those calculated using the interannual DJF data (Fig. 2 of Yu and Lin 2012) in terms of both spatial structure and amplitude. The Ts and Φ500 trends are also broadly similar in the two reanalyses. However, the trends are generally smoother and slightly weaker in 20C compared to NCEP-1, especially for the surface temperature. In addition, modest discrepancies of the Ts trend are seen between the two reanalyses in southwest Asia over 1961–90 and some patches over the United States and Eurasia during 1991–2007.
For 1961–90, the Ts and Φ500 trends in both reanalyses (Fig. 4, left) bear resemblance to the MCA1 pattern (Fig. 3, top left). Thus, it is apparent both visually and from the spatial correlation (Table 2) that the patterns of Ts and Φ500 trends estimated using the decadal filtered data (Trd0) resemble the corresponding trends reconstructed using the MCA1 mode (Trd1) (Fig. 4, center). The spatial correlation of Ts between the two trends is 0.84 (0.89) for NCEP-1 (20C), suggesting that the two trends share about 70% (80%) spatial variances. The pattern correlation between the corresponding Φ500 trends is over 0.90 for both reanalyses. In addition, the pattern correlations for both Ts and Φ500 trends improve slightly when using the trends reconstructed from the two leading MCA modes (Trd2, cf. right panels with left panels in Fig. 4 and in Table 2).
The similarity of the Ts and Φ500 trends estimated using the three approaches is also evident for the period 1991–2007 (Fig. 5). The pattern correlations of the Ts and Φ500 trends between Trd0 and Trd1 are about 0.80 and 0.90, respectively (Table 2). Nevertheless, the reconstructed Ts trend differs from the Trd0 over western central Canada and in patches over southwest Asia, especially for the results in 20C (Fig. 5). The trend difference is also clear in Φ500 over western Canada. The pattern correlations for both temperature and geopotential trends increase slightly using the trends reconstructed from the two leading MCA modes (Table 2).
The temperature and circulation trend analysis indicates that the interdecadal-scale trend change around 1990, as reported in previous studies (e.g., Yu and Lin 2012; Cohen et al. 2012), tends to be a decadal-scale shift in winter and has significant features as characterized by the leading mode of the covariability of decadal land surface temperature and 500-hPa geopotential height anomalies over the northern extratropics. Thus, the widespread winter cooling across the eastern United States and northern Eurasia in the last two decades may be viewed as a manifestation of decadal climate variability internal to the climate system and should not be interpreted as an indication of a slowdown in global warming.
c. Synoptic eddy feedbacks on the circulation and temperature anomalies
The two leading modes of the decadal covariability of temperature and circulation anomalies are characterized by an equivalent barotropic structure in the troposphere. Similar geopotential anomalies are apparent at 250 and 500 hPa in association with the two leading MCA modes (cf. shading in the top panels of Fig. 6 with contours in the top panels of Fig. 3), and similar temperature anomalies are seen at the surface and 850 hPa (cf. shading in the bottom panels of Fig. 6 with shading in the top panels of Fig. 3).
Regressions of (top) 250-hPa geopotential height Φ250 (colors, m2 s−2) and eddy vorticity forcing Fυ250 (contour interval 2 × 10−4 m2 s−3) anomalies on the MCA expansion coefficient of Ts and (bottom) 850-hPa temperature T850 (colors, °C) and eddy heat flux forcing Ft850 (contour interval 3 × 10−7°C s−1) anomalies on the MCA expansion coefficient of Ts for NCEP-1: results related to the (left) first and (right) second MCA modes.
Citation: Journal of Climate 27, 2; 10.1175/JCLI-D-13-00266.1
The centers of action of the two leading MCA modes are located near the midlatitude jet stream and storm tracks (e.g., Peixoto and Oort 1992), suggesting that their generation mechanisms may be associated with local processes such as instabilities of the jet or with transient eddies (e.g., Trenberth and Hurrell 1994; Branstator 1995; Kug and Jin 2009; Yu and Lin 2012). The interaction between synoptic eddies and the low-frequency flow is critical to the formation of climate variability, including the climate variability on decadal–interdecadal time scales (e.g., Zhang et al. 1997; Garreaud and Battisti 1999). It is well recognized that synoptic vorticity forcing is crucial to reinforce the anomalous circulation in the upper troposphere (e.g., Lau 1988; Trenberth and Hurrell 1994; Teng et al. 2007; Kug and Jin 2009; Choi et al. 2010). In the lower troposphere, on the other hand, eddy transients interact with the anomalous mean flow and destroy the mean temperature perturbation (e.g., Lau and Holopainen 1984; Lau and Nath 1991; Hurrell and van Loon 1997). Here we examine the influence of synoptic eddies on the upper-tropospheric circulation and lower-tropospheric temperature anomalies in association with the two leading decadal MCA modes, using the NCEP-1 data.
Figure 6 displays 250-hPa geopotential and synoptic eddy vorticity forcing anomalies (top panels) and 850-hPa temperature and eddy heat flux forcing anomalies (bottom panels) regressed upon the two decadal MCA expansion coefficients of Ts for NCEP-1. At 250 hPa, in association with the MCA1 mode, the dominant anticyclonic circulation forcings are seen over the middle latitudes of North America, North Atlantic, and western Europe, which coincide well with positive geopotential anomalies there. On the other hand, dominant cyclonic forcings are apparent over the high latitudes of the North Atlantic and North Pacific, which are collocated with negative geopotential anomalies. The coincidence of the anticyclonic (cyclonic) vorticity forcing with the positive (negative) geopotential anomaly indicates that synoptic eddies are systematically reinforcing and helping to maintain the anomalous circulation in the upper troposphere. The collocation of vorticity forcing and geopotential height anomalies is also evident over Asian midlatitudes, while the anomalous eddy vorticity forcing is weak. This is probably due to the relatively weak eddy activity in these regions (e.g., Peixoto and Oort 1992; Yu and Lin 2012). Mechanisms that support the geopotential anomalies over Asian midlatitudes remain to be explored. Collocation of the anticyclonic (cyclonic) vorticity forcing with positive (negative) geopotential anomaly is also apparent in the MCA2-associated fields, especially for the anomalies over middle to high latitudes of North America and the North Atlantic and over high latitudes of the North Pacific. However, exceptions are also evident over Eurasian midlatitudes.
In contrast, the pronounced temperature anomalies at 850 hPa coincide with eddy heat flux anomalies with opposite sign (Fig. 6, bottom), indicating that synoptic eddies tend to destroy mean temperature perturbations in a diffusive manner. This also implies that advection by the mean flow is offsetting the eddy forcing and is maintaining the temperature perturbation in the lower troposphere, as discussed above (Fig. 3). Thus, the role of transient eddy forcing on the leading decadal covariability modes is mixed. In the upper troposphere the synoptic vorticity fluxes act to reinforce and maintain the anomalous circulation, while in the lower troposphere the heat flux eddies destroy the decadal temperature perturbation. The result is in agreement with the changes in the storm tracks and related synoptic eddy activity associated with decadal climate variations over the North Pacific and North Atlantic (e.g., Trenberth and Hurrell 1994; Hurrell and van Loon 1997).
d. Associated SST anomalies
The leading expansion coefficient of Ts anomalies bears some resemblance to the decadal filtered DJF mean PDO index (Fig. 7, top left), with a temporal correlation coefficient of 0.53 (0.51) for NCEP-1 (20C). Accordingly, the MCA1-associated SST anomalies share some similarities with the PDO pattern (e.g., Mantua et al. 1997; Zhang et al. 1997) with widespread SST variations in the Pacific basin. In particular, the SST anomalies have a comparable variation in amplitude, changing from positive in the tropical eastern Pacific to negative in the Pacific midlatitudes (Fig. 7, middle and bottom left). However, the MCA1-associated SST anomalies over the Kuroshio differ from those in the central North Pacific, and strong SST warming (cooling) anomalies appear in the North Atlantic middle (high) latitudes, which are probably related to the decadal variation of the thermohaline circulation in these regions (e.g., Manabe and Stouffer 1994). The second expansion coefficient of Ts anomalies is correlated with the decadal filtered DJF mean AMO index (Fig. 7, top right), with a correlation coefficient of 0.95 (0.81) for NCEP-1 (20C). Consequently, the MCA2-associated SST anomalies bear resemblance to the AMO pattern (e.g., van Oldenborgh et al. 2009; Blunden and Arndt 2012) with high SST anomalies in the North Atlantic mid- to high latitudes (Fig. 7, middle and bottom right) and SST anomalies that largely vary with the same sign in the tropical and subtropical Pacific.
(top) Decadal MCA expansion coefficients of Ts for the (left) first and (right) second MCA modes (black for NCEP-1 and blue for 20C) and the decadal-filtered (left) PDO (red) and (right) AMO (red) indices; regressions of SST anomalies on the decadal MCA expansion coefficient of Ts for (middle) NCEP-1 and (bottom) 20C: contour interval 0.1°C.
Citation: Journal of Climate 27, 2; 10.1175/JCLI-D-13-00266.1
5. Decadal covariability in the twentieth century and CanESM2 preindustrial simulation
The decadal temperature and circulation covariability has been examined in the last section using the reanalysis datasets for the period 1951–2010. However, the limited time series makes it impossible to make quantitative estimates about the reliability of the covariability and the significance of the serial correlation considered. In this section, the 20C ensemble-mean data for the period 1901–2010 and the climate model simulation of a 996-yr preindustrial control simulation from CanESM2 are employed to further assess the stationarity of the decadal covariability patterns obtained above and to explore the long-term behavior of the decadal covariability. These datasets are processed in a manner similar to that used for the 1951–2010 period to assess the temperature and circulation variability. Specifically, the linear trend over the 110 (996) DJFs is subtracted from the original data at each grid point to remove the secular variation using normalized orthogonal polynomial approximations. The KZ(7, 2) filter is then applied to assess the decadal–interdecadal variability, together with a 3-DJF cutoff period at both ends of the output. We then perform a MCA between the decadal land surface temperature and 500-hPa geopotential anomalies over the northern extratropics over the 104 (990) decadal filtered DJFs.
a. Covariability in the 20C dataset
The two leading MCA modes, obtained from the 104 decadal filtered DJFs, account for 87.8% (Table 3) of the total squared covariance and are well separated from the next most important covariance mode. In addition, the two MCA modes explain comparable fractions of the total decadal Ts variance (34.5% versus 26.1%) and Φ500 variance (32.7% versus 21.3%). The MCA patterns (not shown) are similar to those described above using the 54 decadal filtered DJFs (Fig. 3), especially for the MCA1 pattern. The main difference between the two MCA analyses using 104 and 54 decadal DJFs appears in the MCA2 mode, with broadly similar spatial patterns between the two but modest differences in the anomalous amplitude detail. Thus, to facilitate a quantitative comparison between the results obtained over the two periods, we project the 104 decadal filtered fields onto the combined MCA patterns (Fig. 3).
As in Table 1 but for the results of the 104 decadal filtered DJFs from the 20C dataset and the 990 filtered DJFs from the CanESM2 simulation.
Figure 8 displays the time series of decadal Ts and Φ500 anomalies of 20C projected on the corresponding spatial patterns of the two leading MCA modes (Fig. 3, top). Slight differences can be seen between the projected time series and the corresponding MCA expansion coefficients (Fig. 3, bottom) for the overlapping period, owing to slightly different climatological means used (110 versus 60 yr) in the calculations. A power spectrum analysis indicates that, for both Ts and Φ500, the MCA1-associated time series has a broad spectrum over decadal–interdecadal time scales while the MCA2-related series has a relatively short time-scale power peak around 20 yr (not shown). In association with the MCA1 mode, the Ts projected time series reveals decadal phase shifts around 1920, 1960, 1977, and 2005. The Φ500 projected series exhibits a generally similar evolution, although the shift around 1960 is not as clear and there is a shift in the early 1980s instead of 1977. The phase shifts seen in Ts around 1920, 1980, and 2005 tend to be related to the well-recognized climate regime shifts that occurred in the 1920s, the late 1970s, and the 2000s (e.g., Graham 1994; Minobe 1997; Mantua et al. 1997; Watanabe and Nitta 1999; Deser et al. 2004; Swanson and Tsonis 2009; Swanson et al. 2009; Lo and Hsu 2010), while the climate regime shift in the 1940s, as reported in previous studies, does not appear in the MCA1-associated time series. Generally, similar evolutions are also apparent in the time series of Ts and Φ500 anomalies projected on the corresponding MCA2 patterns, with about 20-yr fluctuations including a phase shift in the 1940s.
Time series of Ts (blue) and Φ500 (red) anomalies for 20C projected on the corresponding MCA spatial patterns (Fig. 3, top) for the (top) first and (bottom) second MCA modes.
Citation: Journal of Climate 27, 2; 10.1175/JCLI-D-13-00266.1
b. Covariability in the CanESM2 simulation
Figure 9 displays the two leading MCA patterns and corresponding expansion coefficients using the 990 decadal filtered DJFs from the CanESM2 preindustrial control simulation. Statistics related to this MCA analysis are comparable to those obtained from 20C (Tables 1 and 3). The two leading MCA modes also explain comparable fractions of total decadal Ts and Φ500 variances in the CanESM2 simulation.
As in Fig. 3, but for the 990 decadal-filtered DJFs from the CanESM2 model output: contour interval of 30 m2 s−2 for Φ500, negative values hatched.
Citation: Journal of Climate 27, 2; 10.1175/JCLI-D-13-00266.1
The MCA1 pattern (Fig. 9, top left) is dominated by mid to high latitude Ts anomalies that are out of phase with those occurring in the North American and Eurasian subtropics. The correspondence between the climate model simulation and the reanalysis (Fig. 3, top left) is rather good, with a spatial pattern correlation coefficient between the two Ts anomalies of 0.69 over land in the northern extratropics. Nevertheless, the CanESM2 results generally have a much smoother spatial distribution and weaker intensity than that from reanalysis, which is as expected owing to the long model simulation. There are also discrepancies in the detailed placement of the anomalous temperature, especially over mid- to high latitudes of North America. The consistency of temperature anomalies with atmospheric circulation anomalies is also apparent in the CanESM2 simulation. Negative Φ500 anomalies dominate over polar regions and extend slightly southward compared to those in the reanalysis, while geopotential anomalies tend to be positive over the midlatitudes of the North Atlantic and western Europe, Asia, and North Pacific. Consequently, cold temperature advection dominates over mid- to high latitudes of North America, supporting the cold Ts anomalies there. In addition, since the preindustrial simulation we analyzed is driven by prescribed nonevolving atmospheric concentrations of well-mixed gases (including CO2) and natural aerosols or their precursors (e.g., Taylor et al. 2012), the connection between temperature and atmospheric circulation anomalies on the decadal–interdecadal time scale may also be viewed as a manifestation of climate variability internal to the climate system.
The signature of the MCA2-associated temperature and circulation anomalies in the CanESM2 simulation is also qualitatively similar to that in the reanalysis (cf. Fig. 9, top right, with Fig. 3, top right). However, there are noteworthy differences: in particular, the Φ500 anomalies are stronger over the North Pacific and North America in CanESM2 compared to the reanalysis. Modeled geopotential anomalies feature a Pacific–North American (PNA) teleconnection (Wallace and Gutzler 1981), as do the PNA-related Ts anomalies over North America. The MCA2 may also be related to the decadal ENSO-like variability, which is expressed by the classical PNA teleconnection pattern (e.g., Zhang et al. 1997).
The MCA expansion coefficients of Ts and Φ500 anomalies are highly correlated for both modes (Fig. 9, bottom, and Table 3). Figure 10 shows the spectra for the two leading modes of the MCA expansion coefficients of Ts anomalies. The MCA1-related Ts variability exhibits significant power peaks at about 15, 25, and 35–50 yr, while the MCA2-related variability has significant power peaks at about 20 and 30 yr. The relatively shorter time scale of the significant power peak seen in the MCA2-associated Ts variability relative to the MCA1-associated variability also qualitatively resembles the result obtained from 20C (Fig. 8).
Spectra of the two decadal MCA expansion coefficients of Ts anomalies for the CanESM2 model result: variance spectrum of the time series (red curve), variance spectrum of AR1 process fitted to the time series (blue curve), and the 95% upper and lower confidence bounds of the spectrum (green curves).
Citation: Journal of Climate 27, 2; 10.1175/JCLI-D-13-00266.1
6. Relations of Ts and Φ500 variability to atmospheric circulation regimes
Anomalous temperature is closely related to anomalous circulation of the atmosphere and ocean. The two leading modes of decadal covariability of temperature and circulation anomalies (Figs. 3 and 9) reveal a combined characteristic of several known atmospheric teleconnections and, thus, invite comparison of decadal temperature and circulation variability with the well-known atmospheric variability regimes, the North Atlantic Oscillation, and variations over the North Pacific and the tropics. The NAO is associated with changes in the surface westerlies across the Atlantic onto Europe. Following Hurrell (1996), we use the North Pacific index (NPI) and the Southern Oscillation index (SOI) to represent atmospheric variations over the North Pacific and tropics, respectively. The NPI is an area-weighted sea level pressure over the North Pacific (30°–65°N, 160°E–140°W). It is directly related to the variation of Aleutian low pressure, which exchanges midlatitude and polar air and, thus, temperature advection over the North Pacific and west coast of North America. The SOI is defined as the difference of normalized sea level pressures between Tahiti and Darwin. The influence of the SOI and the concurrence of the El Niño–Southern Oscillation extend to the midlatitudes through wavelike patterns that change the jet stream and storm track (e.g., Hurrell 1996; Trenberth et al. 1998).
Figure 11 shows the normalized DJF mean NAO index, negative NPI, and negative SOI, together with their corresponding decadal filtered time series over 1951–2010. On the decadal time scale, the NAO index reveals an upward trend from the 1960s to the early 1990s, following by a slightly downward trend, suggesting that the North Atlantic midlatitude westerly and meridional pressure gradient increase from a weak stage in the 1960s to a strong stage in the last two decades (e.g., Hurrell 1996). The negative NPI is generally in the negative phase before the mid-1970s and then becomes positive, indicating that the Aleutian low has deepened since the late 1970s (e.g., Trenberth 1990; Hanawa et al. 1996), with recoveries around 1990 (e.g., Tachibana et al. 1996) and since the mid-2000s. The negative SOI is high from the late 1970s to the late 1990s and low beyond this period, indicating a change of ENSO activity in the 1980s and 1990s as reported in previous studies (e.g., Wang 1995; An and Wang 2000). The correlations between the expansion coefficient of 500-hPa geopotential anomalies of MCA1 (Fig. 3) and the decadal filtered NAO (−SOI, −NPI) index are 0.75 (0.71, 0.40) for NCEP-1 and 0.73 (0.65, 0.36) for 20C over 1951–2010. In contrast, the correlation between the expansion coefficient of Φ500 anomalies of MCA2 and the three decadal-filtered atmospheric circulation indices is low (less than 0.20).
Time series of normalized DJF-mean NAO (red), negative NPI (green), and negative SOI (blue) indices, together with the correspondingly decadal filtered series.
Citation: Journal of Climate 27, 2; 10.1175/JCLI-D-13-00266.1
The influence of atmospheric circulation regimes on the decadal temperature and circulation anomalies can be estimated through linear regressions using the individual indices or through multivariate linear regression using all three indices. Figure 12 displays the correlation coefficients relating the decadal-filtered Ts and Φ500 anomalies with their corresponding anomalies reconstructed using linear multiple regression on all three indices (left panels) and using the three subcomponents of the multiple regression (right three panels). The main contributions of the three atmospheric regimes on the decadal Ts and Φ500 anomalies, obtained from the three subcomponents of the multivariate regression, are similar to the results obtained from regressions using the individual indices (not shown). The main difference between the results from the two regression approaches appears in low correlation areas. In particular, relatively low and positive correlations from the linear regression approach collocate with low correlations in the individual contributions from the multivariate regression (right three panels in Fig. 12) but with either positive or negative value.
Correlation coefficients relating the decadal filtered Ts (colors) and Φ500 (contours, values greater than 0.6 hatched) with their corresponding anomalies reconstructed using the linear multiple regressions on the three indices: (left)–(right) results for the multiple regressions and the subcomponents of the multiple regressions from the contributions of the NAO, the negative NPI, and the negative SOI alone for the (top) NCEP-1 and (bottom) 20C datasets.
Citation: Journal of Climate 27, 2; 10.1175/JCLI-D-13-00266.1
For decadal surface temperature anomalies (color shadings in Fig. 12), the pronounced NAO contributions (with correlations greater than 0.6) mainly appear over Canada, the eastern United States, Mexico, Greenland, the west coast of Europe, North Africa, and northeast Asia. These are also areas with notable NAO-associated surface temperature anomalies on the interannual time scale (e.g., Hurrell 1996). The main NPI contributions to the decadal Ts variability are seen over northwest Canada and Alaska, North Africa, and mid- to high latitudes over Asia. In contrast, correlations linking the decadal Ts anomalies and the SOI reconstructed ones are relatively low. The weak influence of the SOI on decadal temperature anomalies is probably due to the SOI being an interannual dominated variability. Nevertheless, the SOI contributions (with correlations greater than 0.4) can be seen in the middle parts of Asia, North Africa, and over parts of Canada, the United States, and Greenland. The high correlation counterparts are also seen in the decadal Φ500 anomalies (contours in Fig. 12), indicating the relationship between the large-scale circulation and temperature as discussed above. Thus, the combined effects of all three indices contribute substantially to both decadal Ts and Φ500 anomalies over the northern extratropics (Fig. 12, left). Results from the two reanalyses are similar, with only modest differences in the correlation intensity detail.
The multivariate regression analysis also indicates that the NAO is the most important predictor and the SOI the least important predictor for both decadal Ts and Φ500 anomalies over most of the northern extratropics. Exceptions are apparent over the northwest Canada and Alaska, and northwest Asia where the NPI dominates over the NAO as the most important predictor. This result is similar to that seen from the two reanalyses. The NAO has also been found to be the most important predictor for the interannual variability of surface temperatures and cyclone activities over Canada (e.g., Bonsal et al. 2001; Wang et al. 2006).
7. Summary
This study examines the decadal covariability of the northern wintertime land surface temperature and atmospheric circulation anomalies using the NCEP-1 and 20C datasets as well as a long preindustrial simulation from the Canadian Earth System Model. Changes of temperature and circulation trends around 1990 are related to the leading modes of the decadal covariability, and we diagnose the influence of synoptic eddies on the upper-tropospheric circulation and lower-tropospheric temperature anomalies. We also analyze the relationship between the decadal temperature and circulation variability with several well-known atmospheric variability regimes.
Over the period 1951–2010, about 20% of the wintertime interannual variability of both surface air temperature and 500-hPa geopotential height is on decadal–interdecadal time scales, while the ratio is relatively low for Φ500 over polar regions and the North Pacific. Both temperature and circulation variances have good correspondence in the two reanalyses, although the 20C result is slightly smoother compared to the NCEP-1 result attributed to 20C being the mean of 56 ensemble members. Decadal Ts and Φ500 anomalies covary primarily with the NAO variability but they also covary with the NPI variability, while the SOI variability tends to be the least important predictor for both decadal Ts and Φ500 anomalies over the northern extratropics.
Based on the two reanalysis datasets, the covariability of decadal Ts and Φ500 anomalies over the northern extratropics is found to be dominated by two leading MCA modes, which explain comparable fractions of total decadal temperature and geopotential variances. The leading MCA pattern is characterized by temperature anomalies located over most of Canada, the eastern United States, Mexico, and mid- to high latitudes over Eurasia, accompanied by temperature anomalies of opposite sign over the northeast Canadian Arctic, the western United States, Greenland, North Africa, and parts of the Eurasian subtropics. The second MCA pattern is dominated by temperature anomalies over most of North America, Eurasia, and Greenland, accompanied by anomalies of opposite sign over the west coasts of Europe and North Africa and over East Asian midlatitudes. The consistency of atmospheric circulation anomalies with temperature anomalies is apparent in both modes. The role of transient eddy forcing on the leading decadal covariability modes is mixed. In the upper troposphere the synoptic vorticity fluxes act to reinforce the anomalous circulation, while in the lower troposphere heat flux eddies destroy the decadal temperature perturbation. Hence, it is advection by the anomalous mean flow that is offsetting the eddy forcing and maintaining the temperature perturbation in the lower troposphere.
The change of temperature and geopotential trends around 1990 tends to be a decadal-scale shift in winter and has significant features characterized by the leading mode of the decadal covariability over the northern extratropics. Thus, the widespread winter cooling across the eastern United States and northern Eurasia over the last two decades may be viewed as a manifestation of decadal climate variability internal to the climate system and should not be interpreted as an indication of a slowdown in global warming. In addition, the MCA1-associated Ts and Φ500 variability has a broad spectrum over decadal–interdecadal time scales, while the MCA2-related variability has a relatively short time-scale power peak around 20 yr. The MCA1-associated Ts and Φ500 time series reveal phase shifts around 1920, 1960, 1977, and 2005 that tend to be related to the well-recognized climate regime shifts that occurred in the 1920s, the late 1970s, and the 2000s. The climate regime shift in the 1940s, as reported in previous studies, appears in the MCA2-associated Ts and Φ500 variability. Mechanisms that determine the time scale and phase shift remain to be explored.
Results from a 996-yr preindustrial simulation using CanESM2 exhibit broadly similar leading modes of the decadal covariability, although the simulated pattern is generally smoother in spatial distribution and weaker in intensity compared to that from the reanalysis. The MCA1 pattern in the CanESM2 simulation is dominated by mid- to high-latitude Ts anomalies that are out of phase with those occurring in the North American and Eurasian subtropics. In addition, the CanESM2 simulation features a relatively shorter time scale of the significant power peak in the MCA2-associated Ts and Φ500 variability relative to the MCA1-associated variability. The simulated MCA1-related Ts variability reveals significant power peaks around 15, 25, and 35–50 yr, while the MCA2-related variability has significant power peaks around 20 and 30 yr.
Acknowledgments
We greatly appreciate the work of our colleagues at the CCCma in the production of model results analyzed here. We are thankful to G. P. Compo for helping us with the 20C data and to T. Jang and D. Robataille for help in the data processing. Support for the Twentieth Century Reanalysis Project dataset is provided by the U.S. Department of Energy Office of Science Innovative and Novel Computational Impact on Theory and Experiment (DOE INCITE) program and Office of Biological and Environmental Research (BER) and by the National Oceanic and Atmospheric Administration/Climate Program Office. We also thank the anonymous reviewers for their constructive suggestions and comments, which helped to improve the paper.
REFERENCES
Allen, R. J., and C. S. Zender, 2011: Forcing of the Arctic Oscillation by Eurasian snow cover. J. Climate, 24, 6528–6539.
An, S. I., and B. Wang, 2000: Interdecadal change of the structure of ENSO mode and its impact on the ENSO frequency. J. Climate, 13, 2044–2055.
Arora, V. K., and Coauthors, 2011: Carbon emission limits required to satisfy future representative concentration pathways of greenhouse gases. Geophys. Res. Lett., 38, L05805, doi:10.1029/2010GL046270.
Blunden, J., and D. S. Arndt, Eds., 2012: State of the Climate in 2011. Bull. Amer. Meteor. Soc., 93 (Suppl.), S1–S264.
Bonsal, B. R., A. Shabbar, and K. Higuchi, 2001: Impacts of low frequency variability modes on Canadian winter temperature. Int. J. Climatol., 21, 95–108.
Branstator, G., 1995: Organization of storm track anomalies by recurring low-frequency circulation anomalies. J. Atmos. Sci., 52, 207–226.
Bretherton, C. S., C. Smith, and J. M. Wallace, 1992: An intercomparison of methods for finding coupled patterns in climate data. J. Climate, 5, 541–560.
Choi, D.-H., J.-S. Kug, W.-T. Kwon, F.-F. Jin, H.-J. Baek, and S.-K. Min, 2010: Arctic Oscillation responses to greenhouse warming and role of synoptic eddy feedback. J. Geophys. Res., 115, D17103, doi:10.1029/2010JD014160.
Cohen, J., M. Barlow, P. Kushner, and K. Saito, 2007: Stratosphere–troposphere coupling and links with Eurasian land surface variability. J. Climate, 20, 5335–5343.
Cohen, J., J. C. Furtado, B. Mathew, V. A. Alexeev, and J. E. Cherry, 2012: Arctic warming, increasing snow cover and widespread boreal winter cooling. Environ. Res. Lett., 7, 014007, doi:10.1088/1748-9326/7/1/014007.
Compo, G. P., and Coauthors, 2011: The Twentieth Century Reanalysis project. Quart. J. Roy. Meteor. Soc., 137, 1–28, doi:10.1002/qj.776.
Deser, C., A. S. Phillips, and J. W. Hurrell, 2004: Pacific interdecadal climate variability: Linkages between the tropics and North Pacific during boreal winter since 1900. J. Climate, 17, 3109–3124.
Eskridge, R. E., J. Y. Ku, S. T. Rao, P. S. Porter, and I. G. Zurbenko, 1997: Separating different scales of motion in time series of meteorological variables. Bull. Amer. Meteor. Soc., 78, 1473–1483.
Fletcher, C., P. Kushner, and J. Cohen, 2007: Stratospheric control of the extratropical circulation response to surface forcing. Geophys. Res. Lett., 34, L21802, doi:10.1029/2007GL031626.
Garreaud, R. D., and D. S. Battisti, 1999: Interannual ENSO and interdecadal ENSO-like variability in the Southern Hemisphere tropospheric circulation. J. Climate, 12, 2113–2123.
Gent, P. R., F. O. Bryan, G. Danabasoglu, S. C. Doney, W. R. Holland, W. G. Large, and J. C. McWilliams, 1998: The NCAR Climate System Model global ocean component. J. Climate, 11, 1287–1306.
Graham, N. E., 1994: Decadal-scale climate variability in the 1970s and 1980s: Observations and model results. Climate Dyn., 10, 135–162.
Hanawa, K., S. Ishizaki, and Y. Tanimoto, 1996: Examination of the strengthening of the wintertime mid-latitude westerlies over the North Pacific in the 1970s. J. Meteor. Soc. Japan, 74, 715–721.
Hansen, J., R. Ruedy, M. Sato, and K. Lo, 2010: Global surface temperature change. Rev. Geophys., 48, RG4004, doi:10.1029/2010RG000345.
Hare, S. R., and N. J. Mantua, 2000: Empirical evidence for North Pacific regime shifts in 1977 and 1989. Prog. Oceanogr., 47, 103–145.
Hasselmann, K., 1976: Stochastic climate models. Part I: Theory. Tellus, 28, 473–485.
Hildebrand, F. B., 1956: Introduction to Numerical Analysis. McGraw-Hill, 511 pp.
Hurrell, J. W., 1995: Decadal trends in the North Atlantic Oscillation: Regional temperatures and precipitation. Science, 269, 676–679.
Hurrell, J. W., 1996: Influence of variations in extratropical wintertime teleconnections on Northern Hemisphere temperatures. Geophys. Res. Lett., 23, 665–668.
Hurrell, J. W., and H. van Loon, 1997: Decadal variations associated with the North Atlantic Oscillation. Climatic Change, 36, 301–326.
Hurrell, J. W., Y. Kushnir, M. Visbeck, and G. Ottersen, 2003: An overview of the North Atlantic Oscillation. The North Atlantic Oscillation: Climate Significance and Environmental Impact, Geophys. Monogr., Vol. 134, Amer. Geophys. Union, 1–35.
Kistler, R., and Coauthors, 2001: The NCEP–NCAR 50-Year Reanalysis: Monthly means CD-ROM and documentation. Bull. Amer. Meteor. Soc., 82, 247–267.
Kug, J.-S., and F.-F. Jin, 2009: Left-hand rule for synoptic eddy feedback on low-frequency flow. Geophys. Res. Lett., 36, L05709, doi:10.1029/2008GL036435.
Latif, M.,M. Collins, H. Pohlmann, and N. Keenlyside, 2006: A review of predictability studies of Atlantic sector climate on decadal time scales. J. Climate, 19, 5971–5987.
Lau, N.-C., 1988: Variability of the observed midlatitude storm tracks in relation to the low-frequency changes in the circulation pattern. J. Atmos. Sci., 45, 2718–2743.
Lau, N.-C., and E. Holopainen, 1984: Transient eddy forcing of the time-mean flow as identified by geopotential tendencies. J. Atmos. Sci., 41, 313–328.
Lau, N.-C., and M. J. Nath, 1991: Variability of the baroclinic and barotropic transient eddy forcing associated with monthly changes in the midlatitude storm tracks. J. Atmos. Sci., 48, 2589–2613.
Liu, Z., 2012: Dynamics of interdecadal climate variability: A historical perspective. J. Climate, 25, 1963–1995.
Liu, Z., and M. Alexander, 2007: Atmospheric bridge, oceanic tunnel, and global climatic teleconnections. Rev. Geophys., 45, RG2005, doi:10.1029/2005RG000172.
Lo, T.-T., and H.-H. Hsu, 2008: The early 1950s regime shift in temperature in Taiwan and East Asia. Climate Dyn., 31, 449–461.
Lo, T.-T., and H.-H. Hsu, 2010: Change in the dominant decadal patterns and the late 1980s abrupt warming in the extratropical Northern Hemisphere. Atmos. Sci. Lett., 11, 210–215, doi:10.1002/asl.275.
Manabe, S., and R. J. Stouffer, 1994: Multiple-century response of a coupled ocean–atmosphere model to an increase of atmospheric carbon dioxide. J. Climate, 7, 5–23.
Mantua, N. J., and S. R. Hare, 2002: The Pacific decadal oscillation. J. Oceanogr., 58, 35–44.
Mantua, N. J., S. R. Hare, Y. Zhang, J. M. Wallace, and R. C. Francis, 1997: A Pacific interdecadal climate oscillation with impacts on salmon production. Bull. Amer. Meteor. Soc., 78, 1069–1079.
McFarlane, N. A., J. F. Scinocca, M. Lazare, R. Harvey, D. Verseghy, and J. Li, 2005: The CCCma third generation atmospheric general circulation model. CCCma Internal Rep., 25 pp.
Merrifield, M. A., 2011: A shift in western tropical Pacific sea level trends during the 1990s. J. Climate, 24, 4126–4138.
Minobe, S., 1997: A 50–70 year climatic oscillation over the North Pacific and North America. Geophys. Res. Lett., 24, 683–686.
Murakami, M., 1979: Large-scale aspects of deep convective activity over the GATE area. Mon. Wea. Rev., 107, 994–1013.
Namias, J., 1969: Seasonal interactions between the North Pacific Ocean and the atmosphere during the 1960s. Mon. Wea. Rev., 97, 173–192.
Nitta, T., and S. Yamada, 1989: Recent warming of tropical sea surface temperature and its relationship to the Northern Hemisphere circulation. J. Meteor. Soc. Japan, 67, 375–383.
Overland, J. E., J. M. Adams, and N. A. Bond, 1999: Decadal variability of the Aleutian low and its relation to high-latitude circulation. J. Climate, 12, 1542–1548.
Peings, Y., E. Brun, V. Mauvais, and H. Douville, 2013: How stationary is the relationship between Siberian snow and Arctic Oscillation over the 20th century? Geophys. Res. Lett., 40, 183–188, doi:10.1029/2012GL054083.
Peixoto, J., and A. Oort, 1992: The Physics of Climate. American Institute of Physics, 520 pp.
Power, S. B., and R. Colman, 2006: Multi-year predictability in a coupled general circulation model. Climate Dyn., 26, 247–272.
Rayner, N. A., D. E. Parker, E. B. Horton, C. K. Folland, L. V. Alexander, D. P. Rowell, E. C. Kent, and A. Kaplan, 2003: Global analyses of SST, sea ice and night marine air temperature since the late nineteenth century. J. Geophys. Res., 108, 4407, doi:10.1029/2002JD002670.
Rodionov, S. N., and J. E. Overland, 2005: Application of a sequential regime shift detection method to the Bering Sea ecosystem. ICES J. Mar. Sci., 62, 328–332.
Schlesinger, M. E., and N. Ramankutty, 1994: An oscillation in the global climate system of period 65–70 years. Nature, 367, 723–726.
Solomon, S., D. Qin, M. Manning, Z. Chen, M. Marquis, K. Averyt, M. Tignor, and H. L. Miller Jr., Eds., 2007: Climate Change 2007: The Physical Science Basis. Cambridge University Press, 996 pp.
Swanson, K. L., and A. A. Tsonis, 2009: Has the climate recently shifted? Geophys. Res. Lett., 36, L06711, doi:10.1029/2008GL037022.
Swanson, K. L., G. Sugihara, and A. A. Tsonis, 2009: Long-term natural variability and 20th century climate change. Proc. Natl. Acad. Sci. USA, 106, 16 120–16 130.
Tachibana, Y., M. Honda, and K. Takeuchi, 1996: The abrupt decrease of the sea ice over the southern part of the Sea of Okhotsk in 1989 and its relation to the recent weakening of the Aleutian low. J. Meteor. Soc. Japan, 74, 579–584.
Taylor, K. E., R. J. Stouffer, and G. A. Meehl, 2012: An overview of CMIP5 and the experiment design. Bull. Amer. Meteor. Soc., 93, 485–498.
Teng, Q., J. C. Fyfe, and A. H. Monahan, 2007: Northern Hemisphere circulation regimes: Observed, simulated and predicted. Climate Dyn., 28, 867–879.
Trenberth, K. E., 1990: Recent observed interdecadal climate changes in the Northern Hemisphere. Bull. Amer. Meteor. Soc., 71, 988–993.
Trenberth, K. E., and J. W. Hurrell, 1994: Decadal atmosphere-ocean variations in the Pacific. Climate Dyn., 9, 303–319.
Trenberth, K. E., G. Branstator, D. Karoly, A. Kumar, N. C. Lau, and C. Ropelewski, 1998: Progress during TOGA in understanding and modeling global teleconnections associated with tropical sea surface temperatures. J. Geophys. Res., 103, 14 291–14 324.
van Oldenborgh, G. J., L. A. Raa, H. A. Dijkstra, and S. Y. Philip, 2009: Frequency- or amplitude-dependent effects of the Atlantic meridional overturning on the tropical Pacific Ocean. Ocean Sci., 5, 293–301.
von Salzen, K., and Coauthors, 2013: The Canadian Fourth Generation Atmospheric Global Climate Model (CanAM4). Part I: Representation of physical processes. Atmos.–Ocean, 51, 104–125.
von Storch, H., and F. W. Zwiers, 1999: Statistical Analysis in Climate Research. Cambridge University Press, 494 pp.
Wallace, J. M., and D. S. Gutzler, 1981: Teleconnections in the geopotential height field during the Northern Hemisphere winter. Mon. Wea. Rev., 109, 784–812.
Wallace, J. M., and M. L. Blackmon, 1983: Observations of low-frequency atmospheric variability. Large-Scale Dynamical Processes in the Atmosphere, B. J. Hoskins and R. P. Pearce, Eds., Academic Press, 55–94.
Wallace, J. M., C. Smith, and C. S. Bretherton, 1992: Singular value decomposition of wintertime sea surface temperature and 500-mb height anomalies. J. Climate, 5, 561–576.
Wang, B., 1995: Interdecadal changes in El Niño onset in the last four decades. J. Climate, 8, 267–258.
Wang, X. L., H. Wan, and V. R. Swail, 2006: Observed changes in cyclone activity in Canada and their relationships to major circulation regimes. J. Climate, 19, 896–915.
Watanabe, M., and T. Nitta, 1999: Decadal changes in the atmospheric circulation and associated surface climate variations in the Northern Hemisphere winter. J. Climate, 12, 494–510.
Xie, S.-P., and J. A. Carton, 2004: Tropical Atlantic variability: Patterns, mechanisms, and impacts. Earth Climate: The Ocean-Atmosphere Interaction, Geophys. Monogr., Vol. 147, Amer. Geophys. Union, 121–142.
Yu, B., and H. Lin, 2012: Tropical/extratropical forcing on wintertime variability of the extratropical temperature and circulation. Climate Dyn., 40, 1183–1200, doi:10.1007/s00382-012-1367-3.
Zhang, Y., J. M. Wallace, and D. S. Battisti, 1997: ENSO-like interdecadal variability: 1900–93. J. Climate, 10, 1004–1020.