-
Given the above findings, it is interesting to investigate the impacts of the decadal fluctuation of KOE fronts on the North Pacific storm track, especially knowing that the anomalous storm track could further bring forth significant responses of the atmospheric time-mean flow. Several observational studies and a few modeling studies have examined this issue, yet their results show large controversy. Here, we review the current progress by first remarking on some of the common issues concerning observational studies.
-
Observational studies are subject to several kinds of difficulty and uncertainty. (1) First, many observational studies rely on atmospheric reanalysis data, yet currently available datasets often use low-resolution SST data as boundary forcing to their models and can therefore heavily underestimate the oceanic frontal impacts on the atmosphere. Although data assimilation could remedy this issue to some extent, its effectiveness is uncertain. The fidelity of such research results is thus compromised. This issue was first raised by Frankignoul et al. (2011b, hereafter F11) based on the 2.5°-resolution NCEP-NCAR Reanalysis-1 data. Another widely used reanalysis dataset—the ERA-Interim data—underwent several resolution enhancements of the prescribed SST (1°→0.5°→0.05°) (Dee et al., 2011). Since the resolution of the atmospheric model remained 0.75°, the effective SST refinement is actually 1°− 0.75°. This has immense impacts on atmospheric quantities including the storm track (Masunaga et al., 2015; Parfitt et al., 2017; Zhang et al., 2020a). Smirnov et al. (2015) examined the influence of the OEF’s meridional shift on the atmosphere using both NCEP-NCAR (2.5°) and ERA-Interim (0.75°) and found that the atmospheric response detected in ERA-Interim is 40% stronger than in NCEP-NCAR, suggesting the significance of increased SST resolution. However, using the same NCEP-NCAR dataset and the ERA-Interim on a 1.5° grid, Révelard et al. (2016) found a basically similar atmospheric response to KE bimodal variability and therefore concluded that data assimilation can indeed effectively compensate the effects of crude resolution. This argument, however, does not rule out the possibility that the frontal signature in both datasets is underestimated. (2) Second, since the atmospheric response to midlatitude SST anomaly has proven to be very sensitive to the background atmospheric state (Peng et al., 1995, 1997; Ting and Peng, 1995; Peng and Whitaker, 1999; Okajima et al., 2014), studies focusing on different time periods could yield vastly different results, especially under climate regime shift or climate change conditions where background atmospheric circulation and storm track change dramatically (Yin, 2005; Lu et al., 2010; Wu et al., 2011; Gan et al., 2017). Hence, results based on different time periods must be compared with caution. (3) Third, because the ocean and atmosphere are in constant interaction, any observed state is a mixture of forcing and response, as well as external remote influences such as from ENSO (Kelly and Jones, 1996). It is therefore necessary to rely on statistical methods to separate atmospheric response to atmospheric forcing and external signature. Many widely used methods (Frankignoul et al., 2011a; Liu et al., 2012a, b), including the maximum covariance analysis and the lag-correlation analysis, are built upon an assumption about the atmospheric response delay, that is, the time needed for the atmosphere to fully adjust to the SST anomaly (Frankignoul et al., 2011a; F11 hereafter). The atmospheric response delay was estimated from 3 weeks to 4 months (Ferreira and Frankignoul, 2005; Deser et al., 2007; Smirnov et al., 2015), and is commonly taken as 2 months in recent studies (e.g., Révelard et al., 2016). However, this value is purely empirical, and the atmospheric response has been shown to be very sensitive to it (F11; Smirnov et al., 2015). Other methods, such as synchronous correlation or composite analysis, simply cannot separate forcing and response, and their results are thus hard to interpret. External forcing is typically assumed to be from ENSO in the tropical Pacific and can be removed using linear regression models. However, the different ENSO-removal methods could introduce extra differences in the results. External forcing from other regions and processes, such as the Aleutian-Icelandic Seesaw (Honda and Nakamura, 2001; Honda et al., 2001, 2005), could also be at work but are usually ignored. Hence, observational studies must be carefully performed and interpreted.
-
The very few existent observational studies on atmospheric impacts of the KE bimodal variability are summarized in Table 1. These studies used various definitions of the KE mode index. F11, to our knowledge, is the first research in English literature to directly examine the impacts of KE or KEF decadal variability on the atmosphere. They defined the KE index as the leading principal component (PC) of the latitude of the 14°C isotherm at 200 m below the sea surface between 142°−160°E, using the World Ocean Database 2005 spanning 1979−2007. O’Reilly and Czaja (2015) defined the KE mode index using the maximum covariance analysis method, which involves performing a singular value decomposition on the covariance matrix of the SST gradient and geostrophic velocity (or SSH gradient) anomalies to detect a time series maximizing the covariability of both the dynamic and thermodynamic aspects of the KE. The positive phase of the resultant KE index corresponds to the stable mode of KE, and vice versa. Originally, the index was obtained from the AMSR−AVHRR blended SST and AVISO SSH satellite dataset for 2002−2011 and was then projected back to 1992 based solely on SSH. Révelard et al. (2016) used the KE bimodal index of Qiu et al. (2014, hereafter Q14), which is defined as the SSH anomaly averaged in the region 31°−36°N, 140°−165°E, i.e., the region of the southern recirculation gyre south of the KE. This index is in very high correlation with a more sophisticated synthesized index defined as the average of four normalized time series: the latitude of upstream (west of 153°E) KE, its path length reversed, its strength represented by the SSH difference across the axis, and the strength of the southern recirculation gyre. The advantage of the SSH-based index is that it is much easier to calculate. By this definition, a positive KE index signifies a stable KE. Zhang and Luo (2017) made use of the Luo et al. (2016) index, defined as the difference of the average SSH between two 3°-latitude boxes north and south of 35°N in the upstream KE area. This index, termed the KE dipole index, is by definition similar to the KE strength index which is one of the ingredients of the Q14 mode index. It is shown to represent the variability of the upstream KE meanders and mesoscale eddies, with the positive mode associated with the case of small meanders and reduced eddy kinetic energy, i.e., the stable mode of KE. A comparison of the four indices is shown in Fig. 8a.
Reference Index Index area
(°N, °E)Time range Dataset and
resolutionStorm track response Circulation response Frankignoul et al. (2011b) lat[T14200]1 —, 142−160 1980−2008 NCEP 2.5° — Kamchatka high,
KE lowO’Reilly and Czaja (2015)* SVD[SST, SSH] 32−37, 135−155 1992−2011 ERA-Interim 0.75° west +, east − quadruple over NP Révelard et al. (2016) SSH 31−36, 140−165 1979−2012 ERA-Interim 1.5° downstream + NP high, Alaska low Zhang and Luo (2017)*† SSH DIFF 32−35/35−38,
141−1531993−2015 NCEP 2.5° − ↑ downstream jet ↑ Note: T14=14°C isotherm; □200 = at 200 m depth; □1 = PC1; lat = latitude; ↑ = northward shift; + = strengthening; − = weakening; — = not shown. * indicates that the study used synchronous correlation or composite analysis. † indicates that the study did not remove external forcing from ENSO or prove it small. Zhang and Luo (2017) used the SSH difference between areas south and north of 35°N, which are shown separately in the third column. Table 1. Observational studies on KE bimodal variability impacts on the atmosphere
Figure 8. (a) Time series of KE indices from various literature studies indicated in the legend. The vertical line denotes June 2002, before which the O’Reilly and Czaja (2015) index is projected backward using SSH data. (b) Cross-correlation between the literature KE indices and the Q14 index, with the full (projected) and unprojected indices of O’Reilly and Czaja (2015) shown separately. (c) Time series of OE indices from various literature studies indicated in the legend. References: T12: Taguchi et al. (2012); O18: Okajima et al. (2018); W18: Wills and Thompson (2018); YA19: Yao et al. (2019); YU18: Yuan and Xiao (2018). Here, the Taguchi et al. (2012) index is shown as the winter (DJF) mean of the authors’ monthly indices. Yuan and Xiao (2018) provided indices for each season. Shown here is their winter index. (d) Cross-correlation of literature OE indices with the Q14 KE index. Indices are digitized from the respective references and then normalized about their respective mean and low-pass filtered with a cutoff period of one year.
The atmospheric impacts of the KE decadal variability found in literature studies, as indicated by Table 1, are vastly disparate. Révelard et al. (2016) speculated that the reason their results differ from those of F11 may be because the Q14 index they used is a synthesized index having variability of its four ingredients blended, whereas the F11 index solely represents the meridional shift of the temperature front. In fact, this argument is also relevant when comparing the Révelard et al. (2016) results with those of Zhang and Luo (2017). However, the effectiveness of this synthesized-versus-solo-index argument to explain the different atmospheric impacts is arguably limited, noticing that the Q14 synthesized index bares a high correlation with each of its ingredients (not shown), and with the KE strength index of Zhang and Luo (2017; Figs. 8a, b). The KE index of F11, in fact, is only moderately correlated with that of Q14 during their overlapping period (Fig. 8b). A more valid reason could be that the F11 index is just the leading PC, instead of full variability, and that it is defined by 200 m temperature, a relatively shallow level to define the deep-reaching KE jet and may not be fully representative of its position. Previous studies on the KE dynamics have commonly chosen a deeper level, e.g., the 12°C isotherm at 300 m used by the pioneering work of Mizuno and White (Mizuno and White, 1983). The index of F11 exhibits smaller meridional migration than the SSH-based KE position indices. Further, Révelard et al. (2016) noticed that the original un-projected bi-factor (SST and SSH) index of O’Reilly and Czaja (2015) is very similar to the Q14 index, yet the similarity is lost after the index is projected back to 1992 based on SSH data alone. Révelard et al. (2016) verified that using only the un-projected index results in similar atmospheric impacts with the Q14 index, yet the time series is too short to allow for statistical significance. We notice, however, that while the correlation between the projected index and the Q14 index is indeed lower than the un-projected (0.9), it is still as high as 0.8 (Fig. 8b). Therefore, whether this subtle difference is actually responsible for the contradicting atmospheric impacts is doubtful. Last, as stated above, the crude resolution of atmospheric reanalysis data and the synchronous composite method employed by O’Reilly and Czaja (2015) and Zhang and Luo (2017) hinder confident interpretations of the results. The disparateness of the atmospheric response to these KE indices again suggests the need for further examination.
-
Table 2 summarizes current literature studies on the atmospheric response to decadal variations of the OEF or SAFZ SST. These studies have used a variety of OEF/SAFZ indices (Fig. 8c), therefore Table 2 organizes the references into groups according to the employed index, namely: (1) the latitude of maximum SST gradient, or its leading PC, (2) the mean SST, (3) the mean SST gradient, and (4) the maximum SST gradient. The indices are defined either over the SAFZ (Frankignoul et al., 2011b; Taguchi et al., 2012) or include also the KOC and KE regions (other references in Table 2), but are commonly (imprecisely) addressed as the SAFZ. Nevertheless, since the strongest SST gradient is located in the SAFZ, the inclusion of the KOC and KE regions would presumably not do too much harm. In fact, the coarse resolution of the SST data, which cannot resolve the sharp fronts very well, causes the area with the strongest SST gradient (i.e., the SAFZ) to diffuse to the whole KOE region. The definition of the SAFZ, as is obvious here, is vague. Indeed, albeit different, the indices positively correlate to each other (not shown), and they all exhibit some moderate negative correlation with the Q14 KE index when the latter leads by 2−4 years, whereas the synchronous correlation is low (Fig. 8d). We note here that according to Qiu et al. (2017), the positive KE anomaly should be associated with a concurrent positive OEF-E but negative OEF-W at a delay of 2.5 years (see section 2.2), thus the delayed negative correlation and weak concurrent correlation found here is indicative of the dominance of the OEF-W over the OEF-E in terms of their connection with the KE.
Reference Index Index area (°E, °N) Time range Dataset and resolution Storm track
responseCirculation response Frankignoul et al. (2011b) $ {lat\left[max\left(\nabla T\right)\right]}_{1} $ 145−170, 38−45 1982−2008 OISST 0.25°,
NCEP 2.5°— NPO-WP Yao et al. (2018)*# $ \mathrm{l}\mathrm{a}\mathrm{t}\left[\mathrm{m}\mathrm{a}\mathrm{x}\left(\nabla T\right)\right] $ 145−175, 35−47 1911−2010 HadISST 1°,
20CRv2 2°↑ — Taguchi et al. (2012) $ \overline{T} $ 147.5−165.5, 37.5−42.5 1959−2006 ICOADS 2°,
NCEP 2.5°↑ –PNA Okajima et al. (2018)* $ \overline{T} $ 142−184, 35(36)−42(50) 1958−2010 ICOADS 1°,
JRA55 1.25°low-level ↑,
upper-level −–PNA Wills and Thompson (2018) $ \overline{T} $ 140−171, 36−42 1979−2013 ERA-Interim 1.5°,
ERA-Interim 1.5°— KOE low Yao et al. (2019)* $ \overline{\nabla T} $ 145−175, 35−47 1982−2011 OISST 0.25°,
ERA-Interim 0.75°downstream + — Yuan and Xiao (2018)* $ \overline{\nabla T} $ 140−180, 35−45 1949−2014 HadISST 1°,
NCEP 2.5°— +PNA Yao et al. (2018)*# $ \mathrm{m}\mathrm{a}\mathrm{x}\left(\nabla T\right) $ 145−170, 35−47 1911−2010 HadISST 1°,
20CRv2 2°+ — Note: T = SST; $ \nabla $ = gradient; $\overline \Box $ = mean; lat = latitude; □1 = PC1; ↑ = northward shift; + = strengthening; − = weakening; — = not shown. * indicates that the study used synchronous correlation or composite analysis. # denotes that the paper did not provide a time series of the OEF/SAFZ index. Okajima et al. (2018) used a trapezoid area, and the latitude in (out of) the parentheses indicates the latitude range of the east (west) end of the trapezoid. Studies with a time range covering only 1979 and onwards are denoted with italic fonts. Resolutions before and after the comma indicate SST and atmospheric reanalysis data, respectively. The NPO-WP and PNA patterns referred to in the table are shown in Fig. 9. Table 2. Observational studies on the OEF/SAFZ impacts on the atmosphere
The two references in group 1, respectively, presented results of the storm track and mean-flow responses and are thus hard to compare. Group 2 shows some continuity, only with Wills and Thompson (2018) as an exception. The Taguchi et al. (2012) result is supported by the same authors’ high-resolution (0.5°) coupled model result, thus having added robustness. Group 3 has the same difficulty as group 1, while group 4 has only one reference. Comparing the different groups, since the indices are positively correlated, it is presumed that the groups should have consistent atmospheric impacts, yet this is not the case as shown in Table 2, with the storm track and circulation responses both showing large diversity in terms of both pattern and sign. The NPO-WP and the PNA patterns referred to in the table are shown in Fig. 9. No conclusion could be made based on these results. Taguchi et al. (2012) hypothesized that the discrepancy between their result and the results of F11 might be due to different time range and therefore different climatological background states, with their in-situ data covering 1959−2006, whereas the satellite data used in F11 cover only 1982−2008, the years after the 1977 climate shift. This is an educated guess, considering the large sensitivity to the background state revealed by a number of studies on atmospheric response to midlatitude SST anomalies (see Kushnir et al., 2002; Zhou, 2019), but more examination is needed. Another issue is still the inadequate atmospheric data resolution and the incapable analysis methods.
Figure 9. (a) The positive NPO-WP pattern, defined as the 2nd empirical orthogonal function (EOF) of sea level pressure anomaly (hPa). (b) The positive PNA pattern, defined as the 1st EOF of 500 hPa geopotential height anomaly (m). Anomaly is defined as the monthly deviation from the multi-year mean annual cycle. Based on the NCEP-NCAR Reanalysis 1 dataset for 1948−2018.
-
There are only a few modeling studies on the issue of atmospheric impacts of OEF or SAFZ SST variability (Table 3). Taguchi et al. (2012) analyzed the output of the CFES coupled model (atmospheric resolution 110 km, oceanic resolution 0.5°) and examined the atmospheric response to an increased SAFZ SST. The model results confirmed their findings using observational data, i.e., SAFZ warming leads to a northward shift of the storm track and a negative PNA-like atmospheric response. Later, Okajima et al. (2014) took the atmospheric component of CFES, the AFES, and performed sensitivity experiments regarding the impacts of SAFZ warming in October. Their results showed that the atmospheric response is an equivalent barotropic high over the KOE, driven by a poleward-shifted storm track. Smirnov et al. (2015) forced the CAM5 AGCM with the SST anomaly regressed onto the F11 OEF index, using two different resolutions. The high-resolution (0.25°) run simulates a northward shift of the storm track, accomplished with a high over the Gulf of Alaska and a low over coastal California, which does not agree with the F11 observational results. The atmospheric response in the low-resolution (1°) model, in contrast, is an equivalent barotropic linear response with low-level heating balanced by cold advection (see section 3), as a result of inadequately simulated transient eddy vorticity forcing on the mean flow. The diabatic heating profile is almost identical between the two resolutions, suggesting the important role of resolution in allowing the diabatic heating to be balanced by the right atmospheric process, either transient eddies or the mean flow. Vertical motion in the high-resolution case is much stronger, due to finer-scale structures of diabatic heating in the lower troposphere, and the anomalous mean horizontal flow in the upper levels. More recently, Okajima et al. (2018) again used the 110-km-resolution AFES to study the atmospheric response to SAFZ warming, this time focusing on January and used an artificially inflated SST anomaly. The low-level storm track, they found, shifts poleward and the upper-level storm track weakens downstream. The corresponding circulation response exhibits the negative PNA pattern, similar to, yet weaker than, the observed result of the same authors. Energy budget analysis showed that the main energy source for the circulation response is the background mean-flow available potential energy, while mean-flow and eddy kinetic energies also make remarkable contributions. The reason for the weaker-than-observed atmospheric response was attributed to insufficient conversion from eddy kinetic energy, which was further linked to a weaker background storm track, confirming that the storm track response (and thus the mean flow response) is sensitive to the storm track background state (Peng and Whitaker, 1999; Walter et al., 2001; Brayshaw et al., 2008). In contradiction with the 1°-resolution experiment of Smirnov et al. (2015) who obtained only a linear response, Okajima et al. (2018) used a similar resolution but found a significant eddy-mediated response, presumably due to the artificially inflated SST anomaly compensating the deficiency of low-resolution models in simulating latent heating induced by mesoscale precipitating systems. This argument does not agree with the finding of Smirnov et al. (2015) about nearly identical latent heating across resolutions, probably a result of the different model physics. The atmospheric circulation response to SAFZ warming or OEF northward-shift simulated in these modeling studies invariably shows a high over the North Pacific, yet the location of the high varies. However, they generally agree on the storm track response, suggesting a low-level northward shift, and a high-level downstream weakening, which is consistent with some of the observational results.
Reference Resolution Storm track response Circulation response Taguchi et al. (2012) 110 km ↑ –PNA Okajima et al. (2014) 110 km ↑ KOE high Smirnov et al. (2015) 0.25° ↑ Gulf of Alaska high, coastal California low Smirnov et al. (2015) 1° — linear Okajima et al. (2018) 110 km low-level ↑, upper-level downstream − –PNA Note: ↑ = northward shift; − = weakening; — = not shown. The PNA pattern referred to in the table is shown in Fig. 9. Table 3. Modeling studies on OEF/SAFZ impacts on the atmosphere
-
Based on the above discussions, it is clear that since the pioneering work of F11, observational studies on KOE frontal impacts are still insufficient and inconclusive. Probable reasons include the low data resolution, incapability of some statistical methods, and large sensitivity to the background atmospheric state, as well as the lack of knowledge about the KE and OE fronts themselves. Modeling studies focusing on the atmospheric impacts of OE decadal variability have achieved somewhat consistent results, suggesting that the storm track shifts poleward at low levels and weakens downstream at higher altitudes in response to OE strengthening and poleward migration. The role of atmospheric transient eddy feedback and the importance of model resolution are again highlighted.
Inspired by Qiu et al. (2017), here we take the view that the KE and OE are not independent, but dynamically linked. In fact, Fig. 8d evidently indicates that the OE is indeed negatively correlated with KE with a delay of 2−4 years, or positively correlated when it leads by about 2 years. The periodicity of the lead-lag correlation at ~10 years implies the decadal variability of both the KE and OE. This has some useful implications. Ideally, it is assumed that the KE and OE oscillate exactly at 10-year periods, and have a perfect correlation when KE leads by 2.5 years, as schematically illustrated in Fig. 10a (thick curves). At any time, the storm track and the atmospheric time-mean flow would “feel” the baroclinicity and latent heat forcing from both the KE and OE fronts and respond to them simultaneously. Since the KE and OE are not far from each other, passing-by transient eddies may well receive reinforcement from both and therefore their storm track influences could be accumulated. Further, assuming that their influences accumulate linearly and adopt different linear combination coefficients, we can obtain scenarios of the total atmospheric response from this simple linear regression model:
Figure 10. (a) Time series of idealized KE and OE indices (thick curves), both having a period of 10 years, with KE leading OE by 2.5 years; and the time series of the idealized combined atmospheric impacts of the KE and OE with different linear weights (thin curves). (b) Cross-correlation between the idealized KE index (thick black curve) and the idealized OE index (thick red curve) and the combined atmospheric impacts of the KE and OE with different linear weighs (thin curves).
in which
$ {I}_{\mathrm{A}\mathrm{R}} $ ,$ {I}_{\mathrm{K}\mathrm{E}} $ , and$ {I}_{\mathrm{O}\mathrm{E}} $ are indices for the atmospheric response, the KE, and the OE.$ \alpha $ and$ \beta $ are regression coefficients, and$ \varepsilon $ is the random error which is ignored. The time series of the total atmospheric response using different combinations of$ \alpha $ and$ \beta $ are shown in Fig. 10a (thin curves), and their cross-correlation with the KE index is shown in Fig. 10b. Apparently, the more influential the OE, the closer the total response is to it. If the OE influence is significant, the synchronous atmospheric response to KE must not be the fully developed response, and using synchronous regression to find the atmospheric response to KE variability, as many studies did, would result in low or even reversed correlation. Note that here by “synchronous” we ignore the monthly-scale atmospheric response adjustment and focus on interannual to decadal timescales. Results of OE impacts suffer from the same kind of compromised synchronous correlation, yet since in reality the OEF is indeed stronger than the KEF, this effect very likely affects KE results more severely than those of the OE. Hence, examining the KE and OE atmospheric impacts separately may not be appropriate. This calls for a combined analysis of atmospheric response to KE and OE variability. The different and out-of-phase contributions from the OEF-W and OEF-E branches deserve further examination too, which certainly relies on more knowledge on the dynamics of the fronts. -
Many studies have pointed out the seasonality of atmospheric impacts of the KOE fronts (e.g., Taguchi et al., 2009; Nakamura and Miyama, 2014; Yao et al., 2019). During boreal winter, the large heat content stored beneath the midlatitude seasonal thermocline is exposed to the sea surface by increased wind mixing (Alexander and Deser, 1995; Alexander et al., 1999), which together with the cold and dry East Asian winter monsoon blowing from Siberia, brings forth dramatic air-sea temperature contrast. Thus, the ocean releases a large amount of heat to the overlying atmosphere, marking the most intense air-sea interaction season of the year. Frontal influence on the storm track is therefore the strongest in the cold season, which is exactly the reason why many related studies, like this review, emphasize the cold season. More precisely, aside from the seasonal difference, there are even month-to-month differences within the cold season. Taguchi et al. (2012) observed a negative PNA-like atmospheric response to strengthening SAFZ in December (Table 2), but the SAFZ strengthening in January is not associated with a significant atmospheric response. Further analyses attributed the weak January response to the reduced westerly jet in response to the December SAFZ anomaly, which results in a reduction of upward air-sea fluxes that eventually hinder the frontal influence on the atmosphere in January. Modeling studies of Nakamura et al. (2004, 2008) and Sampe et al. (2010) also found the most evident front-storm track relation in mid-winter and the breakup of such a relationship in late winter. They attributed the late-winter breakup to the enhanced subtropical jet which traps the eddies inside its core and thus cannot respond to the ocean fronts. Such phenomenon, however, is not observed in the modeling results of Okajima et al. (2018). Contrary to the common belief of winter-maxima, Nakamura and Yamane (2010) found the opposite in their analysis based on the NCEP-NCAR reanalysis data and proposed a contradicting view that in winter, the baroclinicity contributed by enhanced quasi-stationary planetary waves and the large ocean-continent contrast outweighs the sensible and latent heating associated with midlatitude oceanic fronts, hence frontal influence on the storm track is even weaker than in summer. The sensitivity of the oceanic frontal impact to the atmospheric background state is thus emphasized, just like in the problem of atmospheric response to large-scale SST anomalies (Kushnir et al., 2002; Zhou, 2019). As the background state varies on monthly, seasonal, and interannual timescales, this problem can be extremely complex.
Moreover, atmospheric general circulation and storm track also undergo multi-decadal and long-term changes, especially under global warming conditions (e.g., Hare and Mantua, 2000; Yin, 2005; Iwao et al., 2012; Willison et al., 2015), which would alter the atmospheric response to frontal anomalies. As introduced above in section 6.3, this was already noticed by Taguchi et al. (2012). Révelard et al. (2016) studied the atmospheric response to KE bimodal variability for two different time periods (1979−2012 and 1959−2016) and found different results. This confirms the differential atmospheric response to frontal variability under different background states. The difference may well be related to the so-called climate regime shift around 1977, which is fundamentally a persistent phase reversal of the Pacific Decadal Oscillation (PDO; e.g., Newman et al. (2016)), yet the mechanism is yet to be fully understood. Furthermore, Qiu et al. (2014) revealed that the KE bimodal variability exhibits weaker amplitude and a shorter period before 1977, according to the OFES model (the oceanic component of CFES) hindcast, suggesting a systematic change pre and post 1977 in both the atmospheric state and the ocean. Hence, a closer investigation into the front-storm track interaction under different atmospheric and oceanic background states, particularly its future changes, is of special research interest.
Reference | Index | Index area (°N, °E) | Time range | Dataset and resolution | Storm track response | Circulation response |
Frankignoul et al. (2011b) | lat[T14200]1 | —, 142−160 | 1980−2008 | NCEP 2.5° | — | Kamchatka high, KE low |
O’Reilly and Czaja (2015)* | SVD[SST, SSH] | 32−37, 135−155 | 1992−2011 | ERA-Interim 0.75° | west +, east − | quadruple over NP |
Révelard et al. (2016) | SSH | 31−36, 140−165 | 1979−2012 | ERA-Interim 1.5° | downstream + | NP high, Alaska low |
Zhang and Luo (2017)*† | SSH DIFF | 32−35/35−38, 141−153 | 1993−2015 | NCEP 2.5° | − ↑ | downstream jet ↑ |
Note: T14=14°C isotherm; □200 = at 200 m depth; □1 = PC1; lat = latitude; ↑ = northward shift; + = strengthening; − = weakening; — = not shown. * indicates that the study used synchronous correlation or composite analysis. † indicates that the study did not remove external forcing from ENSO or prove it small. Zhang and Luo (2017) used the SSH difference between areas south and north of 35°N, which are shown separately in the third column. |