Link between the Barents Oscillation and Recent Boreal Winter Cooling over the Asian Midlatitudes

Observations and reanalysis data have shown that the Asian midlatitudes have experienced a large-scale cooling trend in boreal winter since the late 1980s (Figs. 1a-c) (Honda et al., 2009; Cohen et al., 2012; Outten and Esau, 2012; Outten et al., 2013; Wu et al., 2013). This cooling trend is considered to be related to the loss of Arctic sea ice (Petoukhov and Semenov, 2010; Outten and Esau, 2012; Outten et al., 2013; Wu et al., 2013), increases in both high-latitude moisture and Eurasian snow cover (Cohen et al., 2012), and the Arctic Oscillation (AO) (Thompson and Wallace, 1998; Kryzhov and Gorelits, 2015).

Large-scale atmospheric circulations could change with a continuing loss of Arctic sea ice (Overland and Wang, 2010). (Outten and Esau, 2012) suggested that the recent reduction in Arctic sea-ice concentrations could change the meridional temperature gradient, and hence the large-scale atmospheric flow of the Northern Hemisphere. (Wu et al., 2013) found a tripole wind pattern over northern Eurasia, which could lead to winter precipitation and surface temperature anomalies over the mid-to-high latitudes of Asia.

Furthermore, the trend in this tripole wind pattern since the late 1980s correlates significantly with autumn Arctic sea-ice anomalies. (Cohen et al., 2012) suggested that summer and autumn warming trends in the Northern Hemisphere over the last two decades have coincided with increases in both high-latitude moisture and Eurasian snow cover, which dynamically induce large-scale wintertime cooling. Eurasian mean surface air temperature (SAT) anomalies are also highly correlated with the AO index (Thompson and Wallace, 1998). Usually, winters in northern parts of northern Eurasia are warmer and wetter during positive AO phases, and colder and drier during negative AO phases (Kryzhov and Gorelits, 2015). The AO index has exhibited a decreasing trend since the late 1980s, and therefore the recent boreal winter cooling over the midlatitudes of Asia might be related to this decrease in the AO index. However, (Cohen et al., 2012) questioned whether the boreal winter cooling trend is a consequence of internal variability or a response to changes in boundary forcing —— an issue that remains open to debate.

The Barents Oscillation (BO) was first proposed by (Skeie, 2000). It is the second leading mode of the empirical orthogonal function (EOF) of sea level pressure (SLP) anomalies polewards of 30^°N, and its importance is connected with the climate of the Eurasia, the Nordic seas and the Barents Sea (Skeie, 2000; Chen et al., 2013). In this study, we emphasize that the BO is also important for winter SAT anomalies in the Asian midlatitudes, and that the BO could also contribute to the recent boreal winter cooling in this region. During our EOF analysis, we expand the study area from north of 30^°N to north of 20^°N.

To obtain reliable results, five different datasets are used in this study. The primary variables used are the monthly mean surface or near-surface air temperature, surface wind, SLP, and geopotential height from the HadCRUT4 (Morice et al., 2012), National Centers for Environmental Prediction-National Center for Atmospheric Research (NCEP-NCAR) Reanalysis (Kalnay et al., 1996), Twentieth Century Reanalysis V2 (20CR) (Compo et al., 2011), ERA-Interim (Berrisford et al., 2011; Dee et al., 2011), and ERA-20C (Poli et al., 2013) datasets. For the data of HadCRUT4 and ERA-Interim, the period we use is from 1979 to 2015. For the NCEP-NCAR Reanalysis, the data cover from 1948 to 2015. The 20CR and ERA-20C datasets provide more than 100 years of reanalysis data. The horizontal resolutions of HadCRUT4, NCEP-NCAR Reanalysis, 20CR, ERA-Interim, and ERA-20C are 5^°× 5^°, 2.5^°× 2.5^°, 1^°× 1^°, 0.75^°× 0.75^°, and 1^°× 1^°, respectively.

We focus mainly on the trends in boreal winter (December-February) SAT anomalies over the midlatitudes of Asia (40^°-65^°N, 60^°-140^°E). EOF analysis is applied to the monthly mean SLP north of 20^°N to obtain the second leading mode (i.e., the BO) of the variation in SLP, and the methods of linear trend analysis, correlation analysis, and regression analysis are also employed.

Unlike the global mean surface temperature, which exhibits an increasing trend since 1850, the HadCRUT4, NCEP-NCAR Reanalysis, and ERA-Interim datasets all show a large-scale cooling trend in boreal winter for the past 25 years over the Asian midlatitudes (Figs. 1a-c). These three independent datasets share similar spatial patterns, indicating a strong cooling trend located over northern China, Mongolia, Kazakhstan, and southern Russia (red box in Fig. 1). Previous studies have indicated that boreal winter cooling in Eurasia is related to Arctic sea-ice loss in autumn and winter, and to Arctic warming (Petoukhov and Semenov, 2010; Cohen et al., 2012; Outten and Esau, 2012; Wu et al., 2013). Here, we also examine the linear trend of boreal winter SAT during 1979-1990. Different to the cooling trend of 1990-2015, Figs. 1d-f show a significant warming trend in the area of the red box during 1979-90. Satellite observations have indicated that Arctic sea ice has retreated continuously since October 1978 (Comiso et al., 2008); however, the strong cooling trend has only been observed since the late 1980s. Therefore, this might indicate that other factors, such as the AO and BO, in addition to Arctic sea-ice loss, might also affect the winter climate in this region. To study the boreal winter SAT changes from a long-term perspective, SAT anomalies in the NCEP-NCAR Reanalysis (1948-2015), ERA-Interim (1979-2015) and 20CR (1871-2011) datasets over the Asian midlatitudes are illustrated in Figs. 1g and h. The smoothed lines show that boreal winter SAT in this region contains decadal oscillations. The cooling trend since the late 1980s reflects part of these decadal oscillations, and the warming trend during 1979-90 reflects another part of them. In addition to the recent cooling trend, there is also a cooling trend during 1890-1930 (Fig. 1h), for which the spatial pattern of cooling (not shown) is similar to that shown in Figs. 1a-c.

Figure 1.  Linear trend [^°C (10 yr)^-1] of winter surface or near-surface air temperature during (a-c) 1990-2015 and (d-f) 1979-90, and (g, h) winter SAT anomalies (^°C) over the midlatitudes of Asia [red box in (a)]: (a, d) HadCRUT4; (b, e) NCEP-NCAR Reanalysis; (c, f) ERA-Interim. The blue and red lines in (g) are for NCEP-NCAR Reanalysis and ERA-Interim, respectively; (h) is for 20CR. The thick lines in (g, h) are 10-year smoothed running means.

The first two leading EOF modes of the winter mean SLP are the AO and BO, respectively (Skeie, 2000). The BO has a primary center of action located over the Barents region, and is related to the meridional flow over the Nordic seas and sensible-heat loss in the same region (Skeie, 2000). The BO also correlates with Eurasian SAT anomalies after the AO-related SAT variations are removed. Here, we find that the BO has also contributed to recent winter cooling over the Asian midlatitudes.

Figure 2.  The (a, d) first (AO) and (b, e) second (BO) leading EOF modes of winter monthly mean SLP north of 20^°N, and the (c, f) correlation between the winter SAT anomalies over the midlatitudes of Asia and the BO time-varying index: (a-c) 20CR; (d-f) ERA-20C. The red boxes in (b) and (e) are the same as that in Fig. 1.

Figure 3.  Regression map of boreal winter SAT (shading; ^°C) and surface wind (vectors; m s^-1) on normalized BO time-varying index. The red line is the BO pattern. The period for (a) is 1871-2010, and the data are from 20CR. The period for (b) is 1900-2010, and the data are from the ERA-20C.

The first two leading EOF modes of boreal winter mean SLP north of 20^°N from the 20CR and ERA-20C datasets are shown in Fig. 2. The first leading mode (Figs. 2a and d), which is the AO, accounts for 42% and 44% of the variance for the 20CR and ERA-20C datasets, respectively. The AO has a strong annular structure, and when it is in its positive phase, a ring of strong winds circulates around the North Pole. The second leading mode (Figs. 2b and e), which is the BO, accounts for 10% and 11% of the variance for the 20CR and ERA-20C datasets, respectively. There are three main centers of action for EOF2 (Figs. 2b and e): the Barents center, the western Arctic/Atlantic center, and the Pacific center (Skeie, 2000). The BO has a remarkable meridional structure across the Arctic and the Nordic seas, and anomalous northerly flow over the Nordic seas associated with the BO is concurrent with anomalous southwesterly flow over central Siberia, meaning the BO also correlates with Eurasian SAT anomalies (Skeie, 2000).

Figures 2b and e show that the anomalous SLP center (the Barents center), which is located over the Barents region, has strong meridionality over the high latitudes of Asia. This strong meridionality could also regulate the SAT anomalies in the Asian midlatitudes. The correlation coefficient between the BO time-varying index and boreal winter SAT anomalies in the Asian midlatitudes (red box in Figs. 2b and e) is -0.66 for 20CR and -0.46 for ERA-20C (Figs. 2c and f), and both exceed the 0.01 significance level. The dynamic link between the BO and SAT anomalies over the midlatitudes of Asia is illustrated in Fig. 3. This figure shows that when the BO is in its positive phase, there is an anomalously high SLP center located over the Barents region. The surface wind pattern related to this anomalously high SLP center is clockwise, and the wind direction to the east of the center is southerly. This anomalous southerly wind can transport cold air from the high latitudes to the midlatitudes, and therefore the Asian midlatitudes become cooler. This cooling pattern can reach subtropical regions along eastern parts of China and central and southern Asia. It is worth noting that both independent long-term datasets indicate a weak low SLP center at (34^°N, 100^°E). This low SLP center can restrict the cooling to northern parts of Asia, and thus prevent cooling in subtropical regions of China between 80^°E and 100^°E. Therefore, the geopotential height difference between the anomalously high SLP center and weak low SLP center should directly determine the strength of the anomalous clockwise wind and the magnitude of cooling over the Asian midlatitudes. The correlation coefficient between the boreal winter SAT anomalies over the midlatitudes of Asia and the anomalies of the 1000-hPa geopotential height difference between (65^°N, 60^°E) and (34^°N, 100^°E) is -0.83 during 1971-2011 in 20CR, and -0.88 during 1948-2015 in the NCEP-NCAR Reanalysis (Fig. 4). The BO time-varying index in Fig. 2 and the anomalies of the 1000-hPa geopotential height difference between (65^°N, 60^°E) and (34^°N, 100^°E) in Fig. 4 show that they also contain decadal oscillations. The increasing trends of the BO time-varying index and the 1000-hPa geopotential height difference during 1990-2015, which have contributed to the recent boreal winter cooling over the Asian midlatitudes, are also part of the decadal oscillations.

Figure 4.  Correlation between boreal winter SAT anomalies (^°C) and the anomalies of 1000-hPa geopotential height difference (m) between (65^°N, 60^°E) and (34^°N, 100^°E). Data in (a) are from 20CR. Data in (b) are from NCEP-NCAR Reanalysis.

In this study, the recent boreal winter cooling over the midlatitudes of Asia and its link with the BO are investigated based on observations and reanalysis datasets. The results indicate that the recent large-scale cooling trend since the late 1980s over the Asian midlatitudes is part of the decadal oscillations of long-term SAT anomalies, and that the BO and SAT anomalies in that region correlate significantly in boreal winter. When the BO is in its positive phase, the anomalously high SLP center over the Barents region can force a clockwise wind anomaly able to advect cold air from the high latitudes to the midlatitudes of Asia, resulting in anomalous cold conditions in that region. The recent increase in the BO time-varying index means that the BO has a increasing trend. The increasing BO has also contributed to winter cooling over the Asian midlatitudes during 1990-2015.

The winter cooling trend over the midlatitudes of Asia is related to several other factors, such as a decreasing AO, Arctic sea-ice loss, and increases in both high-latitude moisture and Eurasian snow cover. The determination of the most dominant factor for the recent winter cooling over the midlatitudes should be the subject of further study.