Arctic-North Pacific coupled impacts on the late autumn cold in North America

The Pacific decadal oscillation (PDO) is known to bring an anomalously cold (warm) period to southeastern (northwestern) North America during the cold season of its positive phase through a Rossby wave linkage. This study provides evidence that the remote connection between the North Pacific and the downstream temperature over central North America is strengthened by the warm arctic conditions over the Chukchi and East Siberian Sea, especially in the late autumn season. The modulation effect of the Arctic manifests itself as an altered Rossby wave response to a transient vorticity forcing that results from an equatorward storm track shift, which is induced collaboratively by the PDO and the warm Arctic. This observational finding is supported by two independent modeling experiments: (1) an idealized coupled GCM experiment being nudged toward the warm arctic surface condition and (2) a simple stationary wave model experiment forced by transient eddy forcing.


Introduction
As proximate causes for the recurrent extremely cold winters of 2009/10, 2010/11, 2013/14, and 2014/15 over North America, researchers proposed (1) changes in the wave characteristics due to the arctic sea ice loss and the arctic warming (Stroeve et al 2012, Francis and Vavrus 2015, Overland and Wang 2015, Simmonds 2015 and (2) the warming in the tropical Pacific (Palmer 2014, Hartmann 2015. Compared to the arctic and the tropical impacts, the North Pacific has received less attention in association with potential impacts on the extreme weather in North America (NA). In general, the extratropical sea surface temperature (SST) anomalies are regarded as being inefficient in driving atmospheric anomalies (Lau andNath 1994, Kushnir et al 2002). However, recent studies began to find physical role of extratropical SST (Sato et al 2014, Simmonds andGovekar 2014), and it is also suggested that the abnormal SST warming along the subarctic North Pacific and the western coast of NA during the 2013/14 winter played a non-negligible role in the exceptionally cold winter (Lee et al 2015).
The physical mechanism by which the extratropical SST contributes to the occurrence of extremely cold weather is not yet clear. Here, we emphasize the role of the oceanic baroclinicity and its impacts on the overlying atmosphere. Recently, analyses of observed precipitation and high-resolution modeling experiments supported the anchoring effect of the midlatitude SST on the Pacific storm track nearby the oceanic frontal zones, such as the Kuroshio-Oyashio extension (Sampe et al 2010, Frankignoul et al 2011, Sung et al 2014. The oceanic impacts on the high-frequency eddies have important implications for largescale wave activity, since high-frequency eddies act to enhance low-frequency atmospheric variability (Kug and Jin 2009).
This study investigates the physical links among the North Pacific SST variability, the arctic temperature, and the atmospheric circulation that brings cold weather over the NA region during the late autumn Original content from this work may be used under the terms of the Creative Commons Attribution 3.0 licence.
Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI. season when sea ice-related arctic warming is maximum (Screen and Simmonds 2010). Among the large-scale extratropical SST variabilities over the North Pacific, we focus on the Pacific decadal oscillation (PDO) in this study. It is suggested that the state of the PDO fundamentally affects the weather of North America (Bond andHarrison 2000, Sung et al 2014). Since the Aleutian low, the extratropical oceanic front, and an atmospheric wave bridge that originates from the tropics are major contributors to PDO variability (Schneider and Cornuelle 2005), using the PDO index provides the benefit of reflecting both extratropical and tropical variabilities. Although the name of the PDO highlights the decadal nature of the phenomenon's variability, the interannual variability is also pronounced. This study focuses on the interannual variability of the PDO.

Data and model
We analyzed monthly atmospheric reanalysis data obtained from the National Centers for Environmental Prediction/National Center for Atmospheric Research (Kalnay et al 1996). We set the analysis periods of the dataset for the years 1950-2011, focusing on the late autumn season (October-November). The observational PDO index was downloaded from the Joint Institute for the Study of the Atmosphere and Ocean website (http://jisao.washington. edu/pdo).
In order to complement the observational finding about the impact of the warm and the cold Arctic, we performed two kinds of model experiments using the Geophysical Fluid Dynamics Laboratory CM2.1 coupled model (Delworth et al 2006) and a stationary wave model (SWM) (Ting and Yu 1998). In order to isolate the warm and cold arctic impact in the CM2.1, the SST in the high latitudes (north of 65°N) was restored to the extended reconstructed SST v3 historical reconstruction of Smith et al (2008) covering 60 years, which began in 1951. The restoring time scale was a five-day period. Other regions are fully coupled, and five ensemble simulations were carried out with randomly chosen initial conditions (see Kug et al2015 for further details). The PDO index in the model output was calculated from the leading empirical orthogonal function of SST anomalies in the North Pacific basin (120E-110W, 20N-60N). Since the model is run freely except for Arctic region, the PDO evolves according to the internal dynamics of the model. It is known that the CM2.1 well reproduces the PDO pattern (Park et al 2013), and we could also find the similarity between observational and simulated spatial features of the PDO from figure 1. Moreover, in spite of relatively low spatial resolution of CM2.1 (2.5°×2°), we found that it captures observed spatial pattern of mean storm intensity reasonably well (figure not shown). Although the magnitude of storm activity is weaker than that of observation, it is not a critical issue in this study.
The SWM was used to examine the atmospheric wave response to different source regions. It is a fully nonlinear baroclinic model with a dry dynamical core and 14 vertical levels on sigma coordinates, and horizontal resolution is truncated at rhomboidal 30 (R30). The model experiment was designed to have an idealized forcing function corresponding to the transient vorticity, which has a sine-squared functional form (see Liu et al1998 and Schubert et al 2011 for further details). The background state for the SWM is given by the full three-dimensional climatological flow from October to November reanalysis data.

Climatic impacts of the PDO and the warm arctic during the late autumn
The PDO is known to exert various climatic impacts on NA and its adjacent regions during winter. Under the warm phase of the PDO, i.e., a warm anomaly along the western coast of NA and a cold anomaly in the central North Pacific, NA tends to experience anomalously warm (cold) temperatures over the northwest (the southeast) (Mantua andHare 2002, Wang et al 2013).
Associated with the PDO, the surface air temperature (SAT) exhibits a pronounced seasonal evolution as shown in figure 1. In October, the strong anomalous cold is dominant over the central NA region, then it weakens gradually and shifts eastward during the winter. These evolutionary features in the observation are reproduced well in the CM2.1 model ( figure 1(b)), although the intensity of Z300 response is much weaker than the observation. The change in the SAT is accompanied with an overlying atmospheric circulation, which is similar to the Pacific-North America pattern, including an anomalously deep Aleutian low, an enhanced ridge over northwestern North America, and a downstream trough (Wallace and Gutzler 1981). This pattern is known as an amplified climatological stationary wave pattern, which eases southward advance of the arctic cold air masses that bring the cold anomaly to southeastern NA (Leathers and Palecki 1991). In fact, the cold anomaly is attributed to cold advection due to enhanced northerly winds (see supplementary figure 1). Considering the stronger temperature response over NA in October and November seen in figure 1, we focus on the late autumn season in this study.
We examined the climatic impacts of arctic condition on the atmospheric circulation patterns corresponding to the PDO. In order to represent the arctic condition of the Pacific sector, we used the arctic temperature index (ART2 index) suggested by Kug et al (2015), which is defined as the SAT averaged over the East Siberian-Chukchi Sea region (160°E-160°W, 65°N-80°N) (see supplementary figure 2 for relevant circulation features with the ART2 index). We selected the Pacific Arctic sector, because the temperature variabilities over the Atlantic and Pacific sectors are virtually independent each other and the latter is crucial for the NA climate . It is also important to note that the ART2 and the PDO indices are almost independent each other (r=0.02 and 0.14 for observation and CM2.1 output, respectively. The correlations were not significant even if recent trend in the ART2 index is excluded).
When the PDO and the arctic condition are under warm and cold phases, respectively, (PDO>0 and ART2<0), the temperature and the pressure anomalies in NA show a typical +PDO distribution, leading to northwestern warm and southeastern cold temperatures ( figure 2(a)). On the other hand, there is quite a different situation when the arctic condition is warm (ART2>0). The ridge over northwestern NA moves westward so as to span the offshore NA region and the subarctic coast, and the downstream trough moves farther inland ( figure 2(b)). The altered pressure system enhances the cold anomaly in the central NA region. These altered characteristics of the PDO teleconnection by a warm Arctic are found consistently in the model output (figure 2(d)), although the magnitudes of the anomalies are smaller than those observed. The weaker cold anomaly over central NA in figure 2(d) is partly due to a weaker meridional temperature gradient in the climatology of model atmosphere (figure not shown).

Physical understanding of the modulation effect of a warm arctic
In simple terms, the altered PDO teleconnection patterns by the Arctic may be regarded as a superposition of a conventional PDO teleconnection and a  direct atmospheric response to a warm and a cold Arctic. According to Kug et al (2015), the warm arctic surface condition over the East Siberian-Chukchi Sea often accompanies upper-level anticyclonic and downstream cyclonic anomalies over NA during the winter (December-February). In the late autumn, however, there is no pronounced downstream response over NA (supplementary figure 2(a)). It is interesting that in spite of the smaller downstream impact of the warm Arctic during the late autumn, the downstream branch of the PDO response strengthens and results in a colder NA under the warm Arctic condition, as shown in figure 2. In order to clarify these nonlinear rectification impacts, we compared the PDO teleconnection patterns during the warm and the cold arctic years after removing the direct atmospheric responses to the arctic surface conditions (figure 3). This was implemented by calculating linear regressions onto the PDO index for the subdivision of data according to the sign of the ART2 index. Since the linear regression uses deviations from the temporal average of a variable, the regression coefficients solely for cold (warm) arctic years indicate the PDO teleconnection pattern in which the general circulation features of the cold (the warm) arctic years are removed.
Interestingly, the ridge over northwestern NA and the downstream trough show noteworthy differences between figures 3(a) and (b) regarding the shape and the location. The ridge extends westward corresponding to the warm arctic condition, and the center of the downstream trough is shifted westward as well. In figure 3(b), a steeper slope between the ridge and the downstream trough enables the stronger northerly winds to bring colder advection over central NA. In the observations, we see much colder SAT anomalies prevailing over central NA. The model results also reproduce the similar characteristics of the atmospheric circulation (figures 3(d) and (e)). The only difference from the observations is that the magnitudes of the cold anomalies over NA are almost comparable between the warm and the cold arctic conditions in the model. However, the spatial distribution of cold anomalies seems consistent with that of the observations in which the cold anomaly dominates more over the central northern NA region under the warm arctic condition. The warm anomalies over Alaska show more pronounced differences between the cold and the warm arctic conditions and the observations and model outputs are quite consistent. Figures 3(c) and (f), the difference maps between the warm and the cold arctic conditions, show the distinct impacts of the PDO more clearly. These can be interpreted as nonlinear rectification impacts of the warm Arctic on the teleconnection pattern relevant to the +PDO. The consistency between the observations and the model supports that the warm arctic modulation on the midlatitude teleconnection pattern is not simply due to the linear combination of the direct atmospheric responses to arctic surface warming and the PDO. It is possibly due to the underlying nonlinear process.
The altered Rossby wave patterns in figure 3 might be attributed to different tropical Pacific origins. As the SAT anomalies given in figure 3 suggest, however, there is no noteworthy characteristic in the tropical Pacific. While an El Niño-like anomaly seems stronger in the warm arctic regression of the observations, it is not the same in the model output. Rather than suggesting a tropical source, the different responses to the arctic conditions may be related more to the atmospheric internal process in the extratropics. A large portion of the extratropical low-frequency variability is attributed to internal mechanisms such as barotropic instability and transient eddy forcing   (Ting and Peng 1995, Peng and Whitaker 1999, Li 2004, Lim 2015. Therefore, we examined the behaviors of the Pacific storm track in order to understand the physical link between the altered wave responses to the PDO and the arctic condition. Storm track activity over the North Pacific was defined as follows. 2-8 day bandpass filter was applied to meridional wind at 300 hPa so as to leave the transient component only, and its variance was then defined as the storm track activity. It is known that oceanic anomalies corresponding to the PDO change the meridional temperature gradient in the planetary boundary layer across the Pacific and affect the anchorage of the storm track over the North Pacific (Bond andHarrison 2000, Sung et al 2014). In figures 4(a) and (c), we can see such modulation of the storm track activity with respect to the PDO under the cold arctic conditions for the observations and the CM2.1 output, respectively. Compared to the climatological distribution of the North Pacific storm track (shading), the anomalies relevant to +PDO tend to shift equatorward being centered over 160°W, 40°N (contour). Conspicuously, the shift of the storm track accompanied by the +PDO seems farther equatorward when the arctic surface is under warm conditions (figures 4(b) and (d)), although the shift is not as large in the model as in the observation. The weaker response in the model is possibly due to weaker air-sea interaction in the model (Kushnir et al 2002) or a corollary of weaker activity of transient eddy simulated in CM2.1. Nevertheless, the consistent shift of the storm track farther equatorward under the warm arctic conditions in the observations and the model may imply that the warm Arctic modulates the atmospheric baroclinicity so that the PDO-related SST anomaly can lead to greater shift of the storm track.
In fact, it is not new that warming at the surface of the polar region is associated with an equatorward shift of the storm track (Butler et al 2010, Inoue et al 2012. Moreover, a recent study by Yim et al (2015) shows convincing evidence to support the impact of the warm Arctic on the atmospheric baroclinicity, which can be assessed by the meridional movement of the midlatitude jet stream. In the Coupled Model Intercomparison Project Phase 5 simulations, they found large inter-model diversity in the behaviors of the midlatitude jet under greenhouse warming. By extension, it is also identified that intermodel differences are highly dependent on the intensity of the arctic surface warming simulated in each model, i.e., the midlatitude jet tends to shift equatorward in the models with a stronger arctic surface warming tendency.
The impact of the altered storm track on the largescale circulation was tested using the SWM experiments by prescribing transient vorticity forcing corresponding to the different arctic conditions. Figure 4(e) shows responses to idealized vorticity sources given in two selected locations where anomalous storm activities dominate, i.e., 160°W, 40°N (P1, denoted by C in figure 4(d)) and 160°W, 30°N (P2, denoted by W in figure 4(d)), as shown in figures 4(a) and (b), respectively. Interestingly, the response to the transient vorticity forcing located at P1 (shading) resembles the circulation anomalies shown in figures 3(a) and (d), which correspond to the +PDO under the cold arctic condition, while the response to the vorticity source at P2 is similar to those in figures 3(b) and (e), which display westward shifted circulation anomalies under the warm arctic condition. The agreement between the SWM responses to the anomalous vorticity forcing and the observed circulation anomalies implies that the transient vorticity forcing due to the altered storm track can be critical in generating different Rossby wave train patterns. This is in good agreement with Lim (2015) and Franzke and Feldstein (2005), which showed that an anomalous vorticity transient near the midlatitude storm track can be more important than the diabatic heating for generating certain atmospheric teleconnection patterns. These results also lend credence to the arguments that the warm Arctic can lead to anomalous cold over central NA by modulating the atmospheric teleconnection.

Discussion
In the literature, controversial arguments exist about the origin of the recurring cold winters in North America, as mentioned in the introduction. Researchers who emphasize the role of a warm Arctic present the reasons as the weakened meridional difference in thickness, which leads to weaker thermal winds, and the consequential amplification of the midlatitude atmospheric planetary waves Wang 2010, Francis andVavrus 2015). However, it is still debated whether the warm Arctic really causes a wavier Northern Hemispheric circulation (Barnes 2013, Screen et al 2013, Wallace et al 2014. Our results suggest the influence of the reduced meridional thickness gradient on the low-frequency atmospheric wave in association with the transient vorticity forcing. Although the focus of the current study was the late autumn season, the physical process relevant to the atmospheric baroclinicity and the transient vorticity flux could be applied for the growth of low-frequency waves in the winter as well. Recently, Lee et al (2015) have shown model experiments for one of the most severe winters, i.e., the winter of 2013/14 in NA (see supplementary figure  3 which shows distinct cold anomalies over central NA). Through numerical simulations with various lower boundary forcings, they suggested that the anomalous SST in the North Pacific and the reduced sea ice during the 2013/14 winter could intensify internal variabilities, such as North Pacific oscillation pattern (NPO), so as to bring the exceptional cold winter over the NA region. Their point is quite consistent with ours, which ascribes the origin of the cold weather to the North Pacific SST and the warm Arctic, although the underlying process was not noted in their work. In the present study, by extension, a general relationship was addressed in addition to the physical mechanism for bringing cold weather to NA from the perspective of the PDO. It is supposed that the behaviors of the storm track could be also pertinent to the strengthened NPO variability during the 2013/14 winter, as a synthesis of the results from the anomalous SST structure and the arctic condition. We checked a possible contribution of the North Pacific SST and relevant change in storm track to NA cold in winter using North Pacific Gyre oscillation (NPGO) index instead of the PDO index (supplementary figure  4), since the impact of the PDO-related teleconnection weakens during winter as shown in figure 1 (note the similarity between the SST anomalies of the 2013/14 winter and the NPGO as shown in supplementary  figures 3 and 4(a)). In the same manner with late autumn season, the storm track activity exhibits stronger deviation from its climatology under warm arctic condition, and the corresponding SWM model experiments reproduces similar atmospheric response with that of 2013/14 winter (see supplementary 3 and 4(c)). These results imply that the amplified NPO variability possibly corresponds to the abnormal behaviors of the storm track and resultant anomalous transient vorticity under warm arctic condition. We suppose that the different teleconnection links between late autumn and winter seasons could be understood by seasonal evolutions in basic flows and storm track over the North Pacific.
According to Lu et al (2007), an increase in the subtropical static stability, which is caused by the expansion of the Hadley cell, acts to push the baroclinic instability zone poleward. As suggested previously, the baroclinic instability zone is also responsive to the surface boundary conditions, such as the PDO (Bond andHarrison 2000, Sung et al 2014). Although observations and model experiments show consistent results, the mechanism by which the local surface warming in the narrow arctic region changes the midlatitude atmospheric baroclinicity over the vast expanse of the North Pacific remains unclear. Kidston et al (2011) suggested that the latitude of the sea-ice edge influences the mid-latitude jet stream in the Southern Hemisphere. Likewise, sea ice reduction in the East Siberian-Chukchi Sea might affect atmospheric baroclinity during late autumn season, possibly through the reduced meridional temperature gradient. In this regard, we also note that similar impacts could be induced by sea ice variability over Sea of Okhotsk (Honda et al 1999).
Understanding the potential impact of the Arctic on the internal midlatitude processes is a complicated but important problem, since this may hold the key to understanding, not only the cold winter, but also a recent severe drought in California, which is in the downstream zone of the Pacific storm track (Griffin and Anchukaitis 2014). This motivates further study on the interaction between the arctic and the subarctic circulations.