• Adler, R. F., and Coauthors, 2003: The version-2 Global Precipitation Climatology Project (GPCP) monthly precipitation analysis (1979–present). J. Hydrometeor., 4, 11471167, https://doi.org/10.1175/1525-7541(2003)004<1147:TVGPCP>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Ambrizzi, T., B. J. Hoskins, and H.-H. Hsu, 1995: Rossby wave propagation and teleconnection patterns in the austral winter. J. Atmos. Sci., 52, 36613672, https://doi.org/10.1175/1520-0469(1995)052<3661:RWPATP>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Bayasgalan, G., and J.-B. Ahn, 2018: Seasonal prediction of high-resolution temperature at 2-m height over Mongolia during boreal winter using both coupled general circulation model and artificial neural network. Int. J. Climatol., 38, 54185429, https://doi.org/10.1002/joc.5848.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Beverley, J. D., S. J. Woolnough, L. H. Baker, S. J. Johnson, and A. Weisheimer, 2019: The Northern Hemisphere circumglobal teleconnection in a seasonal forecast model and its relationship to European summer forecast skill. Climate Dyn., 52, 37593771, https://doi.org/10.1007/S00382-018-4371-4.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Dawson, A., 2016: eofs: A library for EOF analysis of meteorological, oceanographic, and climate data. J. Open Res. Software, 4, p.e14, https://doi.org/10.5334/JORS.122.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Ding, Q., and B. Wang, 2005: Circumglobal teleconnection in the Northern Hemisphere summer. J. Climate, 18, 34833505, https://doi.org/10.1175/JCLI3473.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Ding, Q., B. Wang, J. M. Wallace, and G. Branstator, 2011: Tropical–extratropical teleconnections in boreal summer: Observed interannual variability. J. Climate, 24, 18781896, https://doi.org/10.1175/2011JCLI3621.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Enomoto, T., B. J. Hoskins, and Y. Matsuda, 2003: The formation mechanism of the Bonin high in August. Quart. J. Roy. Meteor. Soc., 129, 157178, https://doi.org/10.1256/qj.01.211.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Gao, M., J. Yang, D. Gong, P. Shi, Z. Han, and S.-J. Kim, 2019: Footprints of Atlantic multidecadal oscillation in the low-frequency variation of extreme high temperature in the Northern Hemisphere. J. Climate, 32, 791802, https://doi.org/10.1175/JCLI-D-18-0446.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Hersbach, H., and Coauthors, 2019: Global reanalysis: Goodbye ERA-Interim, hello ERA5. ECMWF Newsletter, No. 159, ECMWF, Reading, United Kingdom, 17–24, https://www.ecmwf.int/node/19027.

  • Hollingsworth, A., K. Arpe, M. Tiedtke, M. Capaldo, and H. Savijärvi, 1980: The performance of a medium-range forecast model in winter—Impact of physical parameterizations. Mon. Wea. Rev., 108, 17361773, https://doi.org/10.1175/1520-0493(1980)108<1736:TPOAMR>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Hong, X., and R. Lu, 2016: The meridional displacement of the summer Asian jet, Silk Road pattern, and tropical SST anomalies. J. Climate, 29, 37533766, https://doi.org/10.1175/JCLI-D-15-0541.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Hoskins, B. J., and D. J. Karoly, 1981: The steady linear response of a spherical atmosphere to thermal and orographic forcing. J. Atmos. Sci., 38, 11791196, https://doi.org/10.1175/1520-0469(1981)038<1179:TSLROA>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Hoskins, B. J., and T. Ambrizzi, 1993: Rossby wave propagation on a realistic longitudinally varying flow. J. Atmos. Sci., 50, 16611671, https://doi.org/10.1175/1520-0469(1993)050<1661:RWPOAR>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kim, H.-J., and J.-B. Ahn, 2015: Improvement in prediction of the Arctic Oscillation with a realistic ocean initial condition in a CGCM. J. Climate, 28, 89518967, https://doi.org/10.1175/JCLI-D-14-00457.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kosaka, Y., H. Nakamura, M. Watanabe, and M. Kimoto, 2009: Analysis on the dynamics of a wave-like teleconnection pattern along the summertime Asian jet based on a reanalysis dataset and climate model simulations. J. Meteor. Soc. Japan, 87, 561580, https://doi.org/10.2151/jmsj.87.561.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kosaka, Y., J. S. Chowdary, S.-P. Xie, Y.-M. Min, and J.-Y. Lee, 2012: Limitations of seasonal predictability for summer climate over East Asia and the northwestern Pacific. J. Climate, 25, 75747589, https://doi.org/10.1175/JCLI-D-12-00009.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Lee, J.-Y., B. Wang, Q. Ding, K.-J. Ha, J.-B. Ahn, A. Kumar, B. Stern, and O. Alves, 2011: How predictable is the Northern Hemisphere summer upper-tropospheric circulation? Climate Dyn., 37, 11891203, https://doi.org/10.1007/S00382-010-0909-9.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Li, R. K. K., T. Woollings, C. O’Reilly, and A. A. Scaife, 2020: Effect of the North Pacific tropospheric waveguide on the fidelity of model El Niño teleconnections. J. Climate, 33, 52235237, https://doi.org/10.1175/JCLI-D-19-0156.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • North, G. R., T. L. Bell, R. F. Cahalan, and F. J. Moeng, 1982: Sampling errors in the estimation of empirical orthogonal functions. Mon. Wea. Rev., 110, 699706, https://doi.org/10.1175/1520-0493(1982)110<0699:SEIEO>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • O’Reilly, C. H., T. Woollings, L. Zanna, and A. Weisheimer, 2018: The impact of tropical precipitation on summertime Euro-Atlantic circulation via a circumglobal wave train. J. Climate, 31, 64816504, https://doi.org/10.1175/JCLI-D-17-0451.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Rowell, D. P., C. K. Folland, K. Maskell, and M. N. Ward, 1995: Variability of summer rainfall over tropical North Africa (1906–92): Observations and modelling. Quart. J. Roy. Meteor. Soc., 121, 669704, https://doi.org/10.1002/QJ.49712152311.

    • Search Google Scholar
    • Export Citation
  • Shaman, J., and E. Tziperman, 2007: Summertime ENSO–North African–Asian jet teleconnection and implications for the Indian monsoons. Geophys. Res. Lett., 34, L11702, https://doi.org/10.1029/2006GL029143.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Stephan, C. C., N. P. Klingaman, and A. G. Turner, 2019: A mechanism for the recently increased interdecadal variability of the Silk Road pattern. J. Climate, 32, 717736, https://doi.org/10.1175/JCLI-D-18-0405.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Sun, J., and J.-B. Ahn, 2015: Dynamical seasonal predictability of the Arctic Oscillation using a CGCM. Int. J. Climatol., 35, 13421353, https://doi.org/10.1002/joc.4060.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Webster, P. J., V. O. Magaña, T. N. Palmer, J. Shukla, R. A. Tomas, M. Yanai, and T. Yasunari, 1998: Monsoons: Processes, predictability, and the prospects for prediction. J. Geophys. Res., 103, 14 45114 510, https://doi.org/10.1029/97JC02719.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Yasui, S., and M. Watanabe, 2010: Forcing processes of the summertime circumglobal teleconnection pattern in a dry AGCM. J. Climate, 23, 20932114, https://doi.org/10.1175/2009JCLI3323.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Zhou, F., R. Zhang, and J. Han, 2019: Relationship between the circumglobal teleconnection and Silk Road pattern over Eurasian continent. Sci. Bull., 64, 374376, https://doi.org/10.1016/j.scib.2019.02.014.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • View in gallery

    Upper-level meridional wind anomalies associated with the Silk Road pattern from the reanalysis and the hindcast runs. Shadings show the 200-hPa meridional wind (m s−1) regressed onto the standardized SR time series identified by the EOF method, for (a) the reanalysis, (b)–(f) the ensemble members, and (g) the ensemble mean. As explained in main text, (a), (f), and (g) show the regression against the first PC, and (b)–(e) show the regression against the second PC. Dots denote the 95% confidence level based on a two-sided Wald test with a t distribution. Black contours denote the climatological 200-hPa zonal wind, with positive solid and negative dashed, contoured at 10 m s−1 and the zero contour omitted.

  • View in gallery

    Temporal correlations and spatial correlations between the Silk Road pattern (SR) in the reanalysis and the hindcast runs, and the fractional variance explained by SR. SR is identified by the empirical orthogonal function method. Member M refers to the ensemble mean. The dotted line in (a) shows the 0.6 lower-limit benchmark for differentiating between synoptically useful patterns. The black line in (b) shows the fractional variance explained by the reanalysis SR.

  • View in gallery

    Time series of the Silk Road pattern in the reanalysis and the hindcast runs. (a) Standardized EOF principal components from the reanalysis and the ensemble mean, and with the linear trends superimposed. Orange crosses represent the ensemble members. As explained in main text, it is the first PC for reanalysis, member 5, and the ensemble mean and the second PC for members 1 to 4. The correlation coefficient between reanalysis PC and ensemble-mean PC is given in parentheses in the panel title. (b) As in (a), but based on detrended data.

  • View in gallery

    Potential predictability in the upper-level meridional wind and in the sea surface temperature in the hindcast runs, and the change in potential predictability after the Silk Road pattern is regressed out. (a) Color shadings show the potential predictability calculated from the 200-hPa meridional winds. (b) The difference between the potential predictability in (a) and the potential predictability calculated after SR is regressed out. (c),(d) As in (a) and (b), but calculated from the sea surface temperatures. The gray contours show the (a),(b) 200-hPa meridional wind and (c),(d) SST regressed onto the standardized time series of the ensemble-mean SR. Contour intervals are (a),(b) 1 m s−1 from ± 0.5 m s−1 and (c),(d) 0.1 K from ± 0.05 K, with positive solid and negative dashed.

  • View in gallery

    Sea surface temperature anomalies associated with the Silk Road pattern from the reanalysis and the hindcast runs. Shadings show the SST (K) regressed onto the standardized SR time series from the EOF method for (a) the reanalysis, (b)–(f) the ensemble members, and (g) the ensemble mean. Dots denote the 95% confidence level based on a two-sided Wald test with a t distribution. Black contours denote the 200-hPa meridional wind regression, with positive solid and negative dashed and with an interval of 1 m s−1.

  • View in gallery

    (a)–(c) Anomalous North Atlantic sea surface temperature and (d)–(f) the Niño-3.4 time series and their correlations with the Silk Road pattern. Orange crosses in (a) and (d) represent the ensemble members. The time series have units of K in (a) and (d), and the times series are standardized in (b), (c), (e), and (f). Correlations and p values are displayed in parentheses in the title of each panel.

  • View in gallery

    Rainfall anomalies associated with the Silk Road pattern from the reanalysis and the hindcast runs. Shadings show the rainfall (mm day−1) regressed onto the standardized SR time series from the EOF method, for (a) the reanalysis, (b)–(f) the ensemble members, and (g) the ensemble mean. Dots denote the 95% confidence level based on a two-sided Wald test with a t distribution. Black contours denote the 200-hPa meridional wind regression, with positive solid and negative dashed and with an interval of 1 m s−1.

  • View in gallery

    Correlations between ENSO and the Indian summer monsoon time series for the reanalysis and the hindcast runs. Standardized time series of Niño-3.4 and ISM for (a) the reanalysis, (b)–(f) the ensemble members, and (g) the ensemble mean. Correlations and p values are displayed in parentheses in the title of each panel.

  • View in gallery

    Upper-level zonal wind bias, and the Rossby waveguides in reanalysis and hindcast run. (a) Climatological ensemble-mean 200-hPa zonal wind bias (shading) from the reanalysis climatology (black contours). (b),(c) Shadings show the stationary wavenumber (Ks) calculated from the climatological 200-hPa zonal wind for (b) reanalysis and (c) the ensemble mean. Solid contours denote Ks = 5 and the black dots denote the 200-hPa easterly zonal wind.

  • View in gallery

    Zonal group velocity of zonally elongated stationary Rossby waves. The zonal group velocity (cg; m s−1) is calculated in the limit kl for (a) reanalysis and (b) the ensemble mean.

All Time Past Year Past 30 Days
Abstract Views 512 474 0
Full Text Views 154 140 19
PDF Downloads 160 138 16

Potential Predictability of the Silk Road Pattern and the Role of SST as Inferred from Seasonal Hindcast Experiments of a Coupled Climate Model

View More View Less
  • 1 Institute of Environment, Energy and Sustainability, The Chinese University of Hong Kong, Hong Kong, China
  • | 2 Institute of Environment, Energy and Sustainability, and Earth System Science Programme, The Chinese University of Hong Kong, Hong Kong, China
  • | 3 Institute of Environment, Energy and Sustainability, and Department of Geography and Resource Management, The Chinese University of Hong Kong, Hong Kong, China
  • | 4 APEC Climate Center, Busan, South Korea
  • | 5 Pusan National University, Busan, South Korea
© Get Permissions
Free access

Abstract

The Silk Road pattern (SR) is a leading mode of atmospheric circulation over midlatitude Eurasia in boreal summer. Its temporal phase is known to be unpredictable in many models. Previous studies have not reached a clear consensus on the role of sea surface temperature (SST) associated with SR. By comparing seasonal hindcasts from the Pusan National University (PNU) coupled general circulation model with reanalysis, we investigate if there are any sources of predictability originating from the SST. It was found that the PNU model cannot predict SR temporally. In fact, SR is associated with El Niño–Southern Oscillation (ENSO) in the model hindcasts, in contrast to reanalysis results in which SR is more associated with North Atlantic SST anomalies. The PNU system, however, shows potential predictability in SR associated with tropical Pacific SST. Bias in stationary Rossby waveguides is proposed as an explanation for the SR–ENSO relationship in hindcast runs. Model upper-level wind bias in the North Atlantic results in a less continuous waveguide connecting the North Atlantic to Asia, and may hinder wave propagations induced by North Atlantic SST to trigger SR. On the other hand, model upper-level wind bias in the subtropical western Pacific may favor westward propagation of zonally elongated waves from the ENSO region to trigger SR. This study implies that the role of SST with regard to SR can be substantially changed depending on the fidelity of model upper-level background winds.

Publisher’s Note: This article was revised on 15 October 2020 to add additional information about author Li's support to the Acknowledgments.

Corresponding author: Dr. Chi-Yung Tam, francis.tam@cuhk.edu.hk

Abstract

The Silk Road pattern (SR) is a leading mode of atmospheric circulation over midlatitude Eurasia in boreal summer. Its temporal phase is known to be unpredictable in many models. Previous studies have not reached a clear consensus on the role of sea surface temperature (SST) associated with SR. By comparing seasonal hindcasts from the Pusan National University (PNU) coupled general circulation model with reanalysis, we investigate if there are any sources of predictability originating from the SST. It was found that the PNU model cannot predict SR temporally. In fact, SR is associated with El Niño–Southern Oscillation (ENSO) in the model hindcasts, in contrast to reanalysis results in which SR is more associated with North Atlantic SST anomalies. The PNU system, however, shows potential predictability in SR associated with tropical Pacific SST. Bias in stationary Rossby waveguides is proposed as an explanation for the SR–ENSO relationship in hindcast runs. Model upper-level wind bias in the North Atlantic results in a less continuous waveguide connecting the North Atlantic to Asia, and may hinder wave propagations induced by North Atlantic SST to trigger SR. On the other hand, model upper-level wind bias in the subtropical western Pacific may favor westward propagation of zonally elongated waves from the ENSO region to trigger SR. This study implies that the role of SST with regard to SR can be substantially changed depending on the fidelity of model upper-level background winds.

Publisher’s Note: This article was revised on 15 October 2020 to add additional information about author Li's support to the Acknowledgments.

Corresponding author: Dr. Chi-Yung Tam, francis.tam@cuhk.edu.hk

1. Introduction

East Asian summer rainfall can be affected by atmospheric teleconnections, and one example is the Silk Road pattern (SR). Enomoto et al. (2003) identified SR as the propagation of a stationary Rossby wave along the upper-level Asian jet, modulating the strength of a climatological subtropical anticyclone near Japan termed the Bonin high. A phenomenon closely related to SR is the circumglobal teleconnection pattern (CGT) (Ding and Wang 2005, their Fig. 4b)—the second leading empirical orthogonal function of interannual variability of 200-hPa geopotential heights over the entire Northern Hemisphere. During the positive phase of CGT, the stationary high just upstream of the weakened Bonin high blocks the eastward-propagating troughs, leading to more rainfall to northern China by moisture advection (Ding and Wang 2005, their Fig. 8). CGT also affects the climate in other regions such as Europe (Ding and Wang 2005; Beverley et al. 2019). Zhou et al. (2019) demonstrated the equivalence of SR and CGT over Eurasia, with correlations reaching 0.87 spatially and 0.68 temporally. As this study focuses on Eurasia, we shall make use of both SR and CGT results from the literature.

Ding and Wang (2005) showed an equivalent barotropic structure throughout CGT but a baroclinic structure in one of the CGT centers of action near northwest India (their Figs. 5 and 6). They proposed that the baroclinic structure is the Gill-type response to the Indian summer monsoon (ISM) heating (their Fig. 4d). Ding and Wang (2005, their Fig. 10), and Ding et al. (2011, their Fig. 8) proposed that ISM can force CGT independent of ENSO. As ENSO can modulate ISM (Ding and Wang 2005, their Fig. 11), ENSO can also affect SR indirectly through ISM. In addition to ISM, Ding et al. (2011, their Fig. 10) tested the CGT generation mechanism with forcings with no preferred structures, and found evidence that basic-state instability can shape CGT and explain its prominence.

To further investigate the forcing mechanism of CGT, Yasui and Watanabe (2010) studied the steady-state responses of a linearized dry atmosphere-only model forced with diabatic heating. They found that remote diabatic heating in equatorial Africa and in southeastern North America can excite CGT. Subsequently, the lower-level circulation anomalies associated with CGT intensify the precipitation and hence the diabatic heating over the eastern Mediterranean, which is a more important driver than the African and North American ones. When climatological diabatic heatings are imposed, Yasui and Watanabe (2010) found that CGT can originate just from internal dry dynamics. They also found that by imposing historical diabatic heating, the temporal variations of CGT were better captured (Yasui and Watanabe 2010, their Figs. 6b,d). However, it has not been addressed whether the diabatic heating patterns are predictable. Yasui and Watanabe (2010, their Fig. 14) also found significant anomalies in the observed SST associated with the positive CGT, with colder anomalies in the North Pacific and the North Atlantic. However, the role of the SST with regard to CGT is still unclear.

Using three coupled climate models, Lee et al. (2011) investigated the predictability of the Northern Hemisphere summertime upper-tropospheric circulation. They found that prediction skills in the multimodel ensemble mainly originate from a zonally symmetric component of the circulation, with a lack of skill in the wave-like CGT. Using the leading and second EOFs to represent the realizable potential predictability, they found that in the observed midlatitudes, regions with significant realizable potential predictability coincide with the centers of a wavelike pattern similar to CGT (Lee et al. 2011, their Figs. 5d and 3d). Meanwhile, their multimodel ensemble cannot perfectly predict this observed predictable mode (Lee et al. 2011, their Fig. 5b).

Using another five different coupled climate models, Kosaka et al. (2012) found that they can only capture the SR pattern spatially but not its temporal occurrence, thus limiting the summertime seasonal predictability in East Asia. SSTs similar to those in Yasui and Watanabe (2010) are also found (Kosaka et al. 2012, their Fig. 11a for negative CGT). Kosaka et al. (2012) noted that while the causality of SST and CGT is unresolved, SST may indicate a weak potential predictability of CGT.

In this study, we evaluate SR prediction skill in a coupled climate model not included in either of these two studies by Lee et al. (2011) and Kosaka et al. (2012). The motivation of our study is to further investigate whether the SST associated with SR provides any sources of potential predictability in our ensemble system, and the mechanism of how the signals from the SST can propagate into the East Asian region of SR. In the remaining parts of this study, section 2 describes the data and our analysis methods. SR identification and reproducibility are shown in section 3. Section 4 investigates the SR potential predictability. Relationships between SSTs and SR triggering are presented in section 5. Conclusions and comparison with other studies are provided in section 6.

2. Data and methods

Seasonally averaged data from June–August (JJA) monthly data from 1980 to 2014 are used. For reanalysis we use the ERA5 product (Hersbach et al. 2019). For precipitation, the Global Precipitation Climatology Project (GPCP) version-2 Monthly Precipitation Analysis is used (Adler et al. 2003). Seasonal hindcasts are obtained from the Pusan National University (PNU) coupled general circulation model (Kim and Ahn 2015). The PNU model is a fully coupled atmosphere–ocean–land–sea ice model, and is part of the APEC Climate Centre (APCC) multimodel ensemble (MME) prediction system. Hindcast runs being used are initialized in May each year. The spread of our five ensemble members is obtained by choosing five different dates for the model initialization in May. Further details of initialization are described in Kim and Ahn (2015). The ensemble mean is obtained from simply (without weight) averaging all ensemble members of the relevant variables, such as wind or SST. Previous studies have shown that the PNU model has skill in predicting large-scale wintertime atmospheric circulation such as the Arctic Oscillation (Sun and Ahn 2015; Kim and Ahn 2015). The model has also been used in seasonal prediction of surface temperature in Mongolia (Bayasgalan and Ahn 2018).

Some SR studies have detrended the data (Ding and Wang 2005; Ding et al. 2011) while others have not (Enomoto et al. 2003; Kosaka et al. 2009; Yasui and Watanabe 2010; Lee et al. 2011). Kosaka et al. (2012) also reported that their SR results were qualitatively similar when their analysis was repeated with linearly detrended data. We shall first present our results based on the raw data, and shall comment on the effect on detrending the data. To facilitate comparison of our results to the more commonly known CGT, we define our SR to have the same phase as CGT in Ding and Wang (2005). It is therefore opposite in phase to SR originally defined by Enomoto et al. (2003).

3. Silk Road pattern in reanalysis and the climate model

a. SR identification by EOF method

To obtain the Silk Road pattern from the reanalysis and PNU hindcast runs, we follow Yasui and Watanabe (2010) and perform empirical orthogonal function (EOF) analysis. The 200-hPa meridional wind is used in the region 20°–60°N, 0°–150°E, with cosine latitude weighting, for computing the EOFs. The much larger region over Eurasia for EOF is less sensitive than the small west-central Asia base point used in Ding and Wang (2005) for identifying CGT. As discussed in Yasui and Watanabe (2010), using the meridional wind is more advantageous than using geopotential height in identifying CGT from reanalysis, because using the latter includes both CGT wave pattern and a zonally symmetric pattern related to the developing ENSO anomalies (Ding et al. 2011).

Figure 1 shows the SR patterns from reanalysis and model, obtained by linearly regressing the 200-hPa meridional wind onto the standardized EOF principal components (PCs). In the reanalysis (Fig. 1a), the leading mode resembles SR with centers of action near 50°, 80°, 110°, and 140°E. The second mode gives a wave pattern in quadrature with SR (see Fig. S1a in the online supplemental material), consistent with results by Yasui and Watanabe (2010) and Kosaka et al. (2012). The fraction of variance explained by the leading and second modes are 0.27 and 0.16 respectively, and the two modes are separated from each other according to North’s rule (North et al. 1982).

Fig. 1.
Fig. 1.

Upper-level meridional wind anomalies associated with the Silk Road pattern from the reanalysis and the hindcast runs. Shadings show the 200-hPa meridional wind (m s−1) regressed onto the standardized SR time series identified by the EOF method, for (a) the reanalysis, (b)–(f) the ensemble members, and (g) the ensemble mean. As explained in main text, (a), (f), and (g) show the regression against the first PC, and (b)–(e) show the regression against the second PC. Dots denote the 95% confidence level based on a two-sided Wald test with a t distribution. Black contours denote the climatological 200-hPa zonal wind, with positive solid and negative dashed, contoured at 10 m s−1 and the zero contour omitted.

Citation: Journal of Climate 33, 22; 10.1175/JCLI-D-20-0235.1

Whereas SR emerged as the leading mode in the reanalysis, the leading mode for 200-hPa meridional wind might not correspond to SR in the model environment. To objectively identify whether the leading or the second mode resembles more of SR, we perform pattern correlation with the reanalysis SR, over the EOF analysis region. SR is then identified as the mode with the highest pattern correlation with its reanalysis counterpart. While SR emerged as the leading mode in the reanalysis and in the ensemble mean, only one in the ensemble (member 5) has SR as its EOF1. The rest of the ensemble members have SR as their EOF2. Only two members (members 1 and 3) have their EOF2 well separated from their EOF1 as in the reanalysis, with two members (members 2 and 5) having marginal separations between EOF1 and EOF2, and one member (member 4) with EOF1 and EOF2 not separated. This may be the reason why some members have the second mode more similar to SR than the leading mode. For the ensemble-mean data, the leading and the second EOF modes are well separated. Therefore, Figs. 1a, 1f, and 1g show the regression of meridional wind against the first PC, and Figs. 1b–e show the regression against the second PC.

The upper-level jets in the reanalysis and the hindcast runs are also shown in Fig. 1, represented by the JJA climatological 200-hPa zonal wind. While the strongest signals of SR are aligned along the Asian jet, it is part of CGT with a zonal wavenumber-5 (k = 5) pattern extending into the North Pacific and North America. The hindcast runs develop jet biases in the seasonal hindcasts (Fig. 9a). The relationship between jet bias and wave propagation will be considered in section 5.

The pattern correlations of SR as identified above are shown in Fig. 2a. Values of all pattern correlations as identified using the above method exceed 0.50 including the ensemble mean, with three members exceeding a correlation of 0.75. These are in accordance with the 0.6 lower limit benchmark for differentiating between synoptically useful patterns (Hollingsworth et al. 1980), giving confidence that our method of SR identification is useful. The fractions of variance explained by the climate model SR are all lower than the fraction of variance explained by the reanalysis SR (Fig. 2b), which may be related to some of the ensemble members having SR as the second mode. The two members with the fraction of variance explained closest to the reanalysis are member 5 and the ensemble mean, consistent with these two having SR as their leading mode.

Fig. 2.
Fig. 2.

Temporal correlations and spatial correlations between the Silk Road pattern (SR) in the reanalysis and the hindcast runs, and the fractional variance explained by SR. SR is identified by the empirical orthogonal function method. Member M refers to the ensemble mean. The dotted line in (a) shows the 0.6 lower-limit benchmark for differentiating between synoptically useful patterns. The black line in (b) shows the fractional variance explained by the reanalysis SR.

Citation: Journal of Climate 33, 22; 10.1175/JCLI-D-20-0235.1

The correlations with the reanalysis SR PC are also shown in Fig. 2a. The temporal correlations are only slightly positive and close to zero, meaning the climate model has little skill in predicting the temporal phase of SR. In addition to the correlation between PCs from reanalysis and hindcast runs, intermember correlation is also examined (Fig. S2). Weak intermember correlations are found, indicating mostly different temporal phases of SR across members. However, the ensemble mean exhibits weak but positive correlation with all the members (Fig. S2g), suggesting that there is potential predictability—albeit weak—of SR in the PNU system.

Figure 3a shows the standardized PCs of SR in the reanalysis and in the hindcast runs. As explained above, it is the first PC for reanalysis, member 5, and ensemble mean and the second PC for members 1 to 4. In the reanalysis, SR exhibits a negative trend with more negative phase SR summers happening in the two recent decades. The most negative SR in the last 35 years was found in 2010, probably initiated by a European blocking in that summer (Kosaka et al. 2012). In contrast, the ensemble mean shows a weak positive trend in SR. We shall discuss the trend in relation to the Atlantic multidecadal oscillation (AMO) in section 4.

Fig. 3.
Fig. 3.

Time series of the Silk Road pattern in the reanalysis and the hindcast runs. (a) Standardized EOF principal components from the reanalysis and the ensemble mean, and with the linear trends superimposed. Orange crosses represent the ensemble members. As explained in main text, it is the first PC for reanalysis, member 5, and the ensemble mean and the second PC for members 1 to 4. The correlation coefficient between reanalysis PC and ensemble-mean PC is given in parentheses in the panel title. (b) As in (a), but based on detrended data.

Citation: Journal of Climate 33, 22; 10.1175/JCLI-D-20-0235.1

b. SR identification by base-point method

To check the robustness of our method in identifying SR, we also use the base-point method. We average the 200-hPa meridional wind over a box north of India (35°–45°N, 75°–85°E) to obtain a time series. We then regress the 200-hPa meridional wind onto the standardized time series, and then invert both the regressed pattern and the time series to obtain the same phase of SR as from the EOF. The regressed patterns (Fig. S3) are similar to those identified through EOF, as reflected by the high pattern correlations in Fig. S4b. The time series are also highly correlated with PCs from the EOF analysis (Fig. S4b). Again, the temporal correlation with reanalysis results is rather low (Fig. S4a). For the rest of the study, we shall use the results from the EOF method.

4. Sources of potential predictability

Potential predictability describes the proportion of variance arising from systematic factors rather than random noise, and assesses the extent to which the future condition can be predicted. Following Rowell et al. (1995), Fig. 4a shows the potential predictability calculated from 200-hPa meridional winds in the ensemble system. The potential predictability decreases with latitude. In the midlatitude, regions with relatively higher potential predictability exist over Asia when compared to the zonal mean. It is noteworthy that these signals coincide with centers of SR (see also Yasui and Watanabe 2010, their Fig. 3b). To investigate the role of SR on the potential predictability, we regress out SR using the standardized PC for each ensemble member. The removed SR in each ensemble member is obtained from multiplying the standardized PC with the regressed meridional wind from Fig. 1. The difference between the original potential predictability and this potential predictability without SR is given in Fig. 4b. Reduced predictability is seen in the three SR centers in Asia, suggesting that indeed SR has associated potential predictability.

Fig. 4.
Fig. 4.

Potential predictability in the upper-level meridional wind and in the sea surface temperature in the hindcast runs, and the change in potential predictability after the Silk Road pattern is regressed out. (a) Color shadings show the potential predictability calculated from the 200-hPa meridional winds. (b) The difference between the potential predictability in (a) and the potential predictability calculated after SR is regressed out. (c),(d) As in (a) and (b), but calculated from the sea surface temperatures. The gray contours show the (a),(b) 200-hPa meridional wind and (c),(d) SST regressed onto the standardized time series of the ensemble-mean SR. Contour intervals are (a),(b) 1 m s−1 from ± 0.5 m s−1 and (c),(d) 0.1 K from ± 0.05 K, with positive solid and negative dashed.

Citation: Journal of Climate 33, 22; 10.1175/JCLI-D-20-0235.1

In addition, we calculate the potential predictability using the ensemble system SSTs (Fig. 4c). Similar to the meridional winds, the tropics have the highest potential predictability, and it decreases with latitude. After SR is regressed out, the largest reduction in potential predictability is located in the tropical eastern Pacific (Fig. 4d). This suggests that the source of SR potential predictability may originate from the SSTs in the ENSO region.

Figure 5 shows regression maps of the SST to the standardized SR PCs in both reanalysis and hindcast runs. The positive SR in the reanalysis is associated with negative SST anomalies in the whole North Atlantic, as well as in the western and central regions of the North Pacific (Fig. 5a), resembling negative AMO SST anomalies. Similar signals have been identified in Yasui and Watanabe (2010, their Fig. 14) and Kosaka et al. (2012, their Fig. 11a for negative CGT), and also closely resemble those anomalies in Hong and Lu (2016, their Fig. 5b). Since SR is a wavelike pattern and the Asian jet acts as a waveguide for Rossby wave, one possible mechanism is that the North Atlantic SST anomalies generate anomalous Rossby wave sources and trigger wave propagation from the North Atlantic into Asia. This will be discussed in section 5.

Fig. 5.
Fig. 5.

Sea surface temperature anomalies associated with the Silk Road pattern from the reanalysis and the hindcast runs. Shadings show the SST (K) regressed onto the standardized SR time series from the EOF method for (a) the reanalysis, (b)–(f) the ensemble members, and (g) the ensemble mean. Dots denote the 95% confidence level based on a two-sided Wald test with a t distribution. Black contours denote the 200-hPa meridional wind regression, with positive solid and negative dashed and with an interval of 1 m s−1.

Citation: Journal of Climate 33, 22; 10.1175/JCLI-D-20-0235.1

In contrast to the North Atlantic SST anomalies resembling the negative AMO in the reanalysis, most of the ensemble members and the ensemble mean (Figs. 5b–g) show La Niña–like SST anomalies. Here we compare the time series for the North Atlantic SST and the Niño-3.4 in both the reanalysis and the hindcast experiments. The North Atlantic SST index is calculated by averaging the SSTs in 0°–65°N, 80°W–0°E with latitudinal weighting, whereas the region used for the Niño-3.4 index is 5°S–5°N, 170°–120°W. Figure 6 shows the temporal correlations between the indices of the North Atlantic SST, Niño-3.4, and SR in the reanalysis and the ensemble mean. As the seasonal hindcasts used observed SSTs for initialization, the correlations between the reanalysis and the ensemble mean are strong for both the North Atlantic SST (0.40; Fig. 6a) and the Niño-3.4 (0.60; Fig. 6d). In the reanalysis, SR strongly and negatively correlates with the North Atlantic SST (−0.48; Fig. 6b), but the correlation between SR and the Niño-3.4 is weak (0.07; Fig. 6e). On the other hand, the model ensemble-mean SR strongly and negatively correlates with the Niño-3.4 (−0.41; Fig. 6f), but the correlation between the ensemble-mean SR and the North Atlantic SST is weak (0.12; Fig. 6c).

Fig. 6.
Fig. 6.

(a)–(c) Anomalous North Atlantic sea surface temperature and (d)–(f) the Niño-3.4 time series and their correlations with the Silk Road pattern. Orange crosses in (a) and (d) represent the ensemble members. The time series have units of K in (a) and (d), and the times series are standardized in (b), (c), (e), and (f). Correlations and p values are displayed in parentheses in the title of each panel.

Citation: Journal of Climate 33, 22; 10.1175/JCLI-D-20-0235.1

The North Atlantic SST in the reanalysis transitioned from an overall negative to an overall positive phase in our period of 1980 to 2014 (Fig. 6a), whereas SR in the reanalysis transitioned from an overall positive to an overall negative phase (Fig. 3a). To investigate how much the association between SR and the North Atlantic SST in the reanalysis is due to the decadal variation, we repeat the EOF analysis by first removing the linear trend of the datasets. Figure 3b shows the SR time series obtained from the detrended data. The temporal correlation between the reanalysis and the ensemble-mean SR is close to zero, similar to the result without detrending (Fig. 3a). The spatial pattern of SR is also similar to that without detrending (not shown). The regressed SST shows a weaker amplitude (Fig. S5a), but importantly these residual SST anomalies still closely resemble the SST anomalies without detrending (Fig. 5a). We therefore conclude that the North Atlantic SST anomalies associated with the reanalysis SR include both the decadal signal and an interannual signal.

Figure S6a shows the time series of the interannual North Atlantic SST signal. It again shows a strong correlation between reanalysis and ensemble-mean hindcast results. In the reanalysis, the correlation between the North Atlantic SST and SR (−0.26; Fig. S6b) is weaker than without detrending (−0.48; Fig. 6b), indicating the partial contribution of decadal component of North Atlantic SST to SR. This correlation is still stronger than the correlation between the ensemble-mean North Atlantic SST and the ensemble-mean SR (0.11; Fig. S6c). Meanwhile, the detrending does not change the results much concerning ENSO (cf. Figs. S6e,f with Figs. 6e,f respectively). Therefore, our conclusion based on detrended data is in line with data without detrending, namely that SR is associated more with North Atlantic SST anomalies in the reanalysis, but more with ENSO-like SST anomalies in our hindcast runs.

5. Connections of SSTs to SR by the Indian summer monsoon and by waveguides

a. Indian summer monsoon

To understand the relationship between SR and ENSO in the hindcast runs, we revisit the forcing mechanisms of SR. The Indian summer monsoon (ISM) has been proposed as a forcing for SR (Ding and Wang 2005; Ding et al. 2011; Stephan et al. 2019). Figure 7 shows the rainfall regressed onto SR. GPCP rainfall is used in the regression for reanalysis SR (Fig. 7a). ISM is enhanced during positive phase of SR in reanalysis and is consistent with results of Ding and Wang (2005, their Fig. 4d), especially over northern India. ISM is also enhanced in all hindcast members and the ensemble mean (Figs. 7b–g).

Fig. 7.
Fig. 7.

Rainfall anomalies associated with the Silk Road pattern from the reanalysis and the hindcast runs. Shadings show the rainfall (mm day−1) regressed onto the standardized SR time series from the EOF method, for (a) the reanalysis, (b)–(f) the ensemble members, and (g) the ensemble mean. Dots denote the 95% confidence level based on a two-sided Wald test with a t distribution. Black contours denote the 200-hPa meridional wind regression, with positive solid and negative dashed and with an interval of 1 m s−1.

Citation: Journal of Climate 33, 22; 10.1175/JCLI-D-20-0235.1

In JJA, ENSO has a negative correlation with ISM due to the perturbed Walker circulation (Webster et al. 1998, their Figs. 19b and 9, respectively). Based on Fig. 7a, we define an ISM index as the rainfall averaged over northern India (20°–30°N, 70°–90°E). Figure 8 shows the standardized time series of Niño-3.4 and ISM in reanalysis and hindcast runs. In reanalysis (Fig. 8a), the correlation between ENSO and ISM is −0.42. The imperfect relationship is consistent with reanalysis SR not being associated with ENSO-like SST (Fig. 5a), despite being associated with ISM. In comparison, the ENSO–ISM correlation ranges from −0.33 to −0.66 in the ensemble members, with an averaged correlation of −0.43. It can be concluded that the hindcast runs generally capture well the relationship between ENSO and ISM. As the ensemble mean reduces greatly the atmospheric internal variability but not the ENSO variability due to slower oceanic evolution time scale, it is expected to see a larger ENSO–ISM correlation in the ensemble mean (−0.56) than in reanalysis. To understand the SR–ENSO relationship in the hindcast runs, we need to consider mechanisms other than ISM forcing SR.

Fig. 8.
Fig. 8.

Correlations between ENSO and the Indian summer monsoon time series for the reanalysis and the hindcast runs. Standardized time series of Niño-3.4 and ISM for (a) the reanalysis, (b)–(f) the ensemble members, and (g) the ensemble mean. Correlations and p values are displayed in parentheses in the title of each panel.

Citation: Journal of Climate 33, 22; 10.1175/JCLI-D-20-0235.1

b. Waveguides

The three members with the SR–ENSO relationship also show meridional wind anomalies near the ENSO-like SST, suggestive of Rossby waves linking the equatorial Pacific and the extratropics (members 1, 3, and 4 from Fig. 5). These waves are not seen in members 2 and 5, which do not show the SR–ENSO relationship. To reveal the linkages between possible wave forcings to occurrence of SR, we revert to the theory of atmospheric Rossby wave propagation and explore the waveguides.

Rossby wave propagation is sensitive to the waveguides (Hoskins and Ambrizzi 1993; Li et al. 2020). The summer Asian jet acts as a waveguide for Rossby waves (Ambrizzi et al. 1995). The 200-hPa ensemble-mean climatological zonal wind exhibits large biases near the Asian jet (Fig. 9a). The ensemble mean shows enhanced westerlies over the North Atlantic jet exit and near the Sea of Japan, as well as in the subtropical western Pacific. These biases are similar across individual members (Fig. S7). To understand how these wind biases affect the waveguides, we compute the stationary total wavenumber Ks from the 200-hPa zonal winds. Following Hoskins and Karoly (1981),

Ks=(βMu¯M)1/2=[2Ωacos2ϕy1cos2ϕy(u¯cosϕcos2ϕ)u¯cosϕ]1/2,

where βM is cosϕ times the meridional gradient of the absolute vorticity on the sphere, and u¯M is the basic zonal velocity on the Mercator projection on the sphere. The term Ks is a useful quantity to understand the wave trapping effect of jets (i.e., as waveguides). When Ks shows a local latitudinal maximum (Ks,max) between two minima (Ks,min), such as in the Asian jet, waves with zonal wavenumbers between Ks,max and Ks,min tend to be trapped (Hoskins and Ambrizzi 1993, their Fig. 2e).

Fig. 9.
Fig. 9.

Upper-level zonal wind bias, and the Rossby waveguides in reanalysis and hindcast run. (a) Climatological ensemble-mean 200-hPa zonal wind bias (shading) from the reanalysis climatology (black contours). (b),(c) Shadings show the stationary wavenumber (Ks) calculated from the climatological 200-hPa zonal wind for (b) reanalysis and (c) the ensemble mean. Solid contours denote Ks = 5 and the black dots denote the 200-hPa easterly zonal wind.

Citation: Journal of Climate 33, 22; 10.1175/JCLI-D-20-0235.1

As SR is part of CGT with k = 5, we focus on the Ks = 5 waveguide inside which Ks > 5 regions can also trap k = 5 waves with nonzero l. In the reanalysis (Fig. 9b), the North Atlantic Ks = 5 waveguide exit merges with the Asian waveguide entrance over the Iberian Peninsula (near 40°N, 0°E), resulting in a continuous waveguide to trap k = 5 waves from the North Atlantic to Asia. In the model simulations (Fig. 9c), the enhanced westerlies in the North Atlantic jet exit result in a discontinuation in the Ks = 5 waveguide between the North Atlantic jet and the Asian jet. If model North Atlantic SST can generate RWS and trigger wave propagation in the North Atlantic, the discontinuation to the Asian waveguide entrance may mean that such waves cannot propagate as easily into Asia to trigger SR in the model environment. This is consistent with SR being less associated with North Atlantic SST in the hindcast runs.

The effects of the wind biases near the Sea of Japan and in the subtropical western Pacific on the waveguides are more subtle. Wind bias near the Sea of Japan has little effect on the Ks = 5 waveguide. However, the enhanced westerlies in the subtropical western Pacific alter the climatological easterlies boundary. The model basic state (Fig. 9c) supports wave propagation in the subtropical western Pacific (15°–30°N, 135°–165°E); this feature is not seen in the reanalysis results where the climatological zonal winds are easterlies and waves are evanescent (Fig. 9b).

Rossby waves typically have eastward group velocities, but there are occasions in which westward dispersion is allowed (Hoskins and Ambrizzi 1993; O’Reilly et al. 2018). Following O’Reilly et al. [2018, their Eq. (5)], we compute the zonal group velocities (cg) of stationary Rossby waves in the limit kl,

cg=u¯MβMl2,

where k and l are the zonal and meridional wavenumbers, respectively. We use a meridional half-wavelength of π/l = 3.5 × 106 m, taken from the westward-propagating waves from the Pacific into Asia as simulated in O’Reilly et al. (2018, their Fig. 8c) and in Hoskins and Ambrizzi (1993, their Fig. 11). For these zonally elongated waves with a limited meridional extent, westward group velocities are found along the southern flank of the Asian jet (near 30°–45°N, 60°–180°E) in both reanalysis and hindcast runs (Figs. 10a,b). These westward dispersions are important when we consider the aforementioned model bias in the subtropical western Pacific. The evanescent region between the Asian jet and the subtropical western Pacific is reduced in the hindcast runs (Fig. 9c). Therefore, it is likely that westward Rossby wave dispersion triggered in the subtropical Pacific into Asia is more favored in the model than reanalysis, especially for the long waves allowed by theory, which have larger zonal extent than the evanescent region. This can explain why the model SR is more associated with ENSO-like SSTs than in reanalysis.

Fig. 10.
Fig. 10.

Zonal group velocity of zonally elongated stationary Rossby waves. The zonal group velocity (cg; m s−1) is calculated in the limit kl for (a) reanalysis and (b) the ensemble mean.

Citation: Journal of Climate 33, 22; 10.1175/JCLI-D-20-0235.1

6. Conclusions and discussion

This study aims to infer the role of sea surface temperatures in the Silk Road pattern formation, by comparing reanalysis with seasonal hindcast experiments of a coupled climate model. Although the hindcast runs cannot predict SR temporally (Fig. 3), the ensemble system shows potential predictability in SR related to tropical Pacific SST (Fig. 4). While reanalysis SR is more associated with North Atlantic SST anomalies similar to the AMO pattern, hindcast runs SR is more associated with tropical Pacific SST anomalies similar to ENSO (Figs. 5 and 6). We propose that bias in the basic state on which Rossby wave propagates explains the association of SR with different SSTs. Analysis of the upper level zonal wind reveals two main biases (Fig. 9a). Bias in the North Atlantic jet exit region results in a discontinuous waveguide from the North Atlantic to Asia (Fig. 9c), which may hinder the propagation of waves associated with North Atlantic SST to trigger SR. Westerly bias in the subtropical western Pacific reduces the evanescent region between the subtropical western Pacific and the Asian jet (Fig. 9c), and may favor the westward dispersion of zonally elongated waves associated with ENSO SST to trigger SR (Fig. 10b).

Similar to Yasui and Watanabe (2010) and Kosaka et al. (2012), we find that the reanalysis SR is associated with North Atlantic SST anomalies that resemble the AMO pattern. AMO SST has recently been shown to modulate extreme high temperature events circumglobally in the Northern Hemisphere summer. Gao et al. (2019) identified five extreme high temperature hotspots on the decadal time scale, located at western North America and Mexico, Europe, central Asia, the Mongolian Plateau, and eastern Siberia (Gao et al. 2019, their Fig. 1). With the exception of central Asia, four of these five hotspots show high temporal correlations among them, suggesting a coherent variation in extreme high temperature events in the four hotspots (Gao et al. 2019, their Table 1). Meanwhile, a barotropic wave pattern associated with the positive AMO exhibits five centers of higher geopotential heights, corresponding to the above four coherent extreme high temperature hotspots, with the remaining center being at southern Greenland (Gao et al. 2019, their Fig. 5d). These five regions of higher heights in the positive AMO also closely resemble the centers of the negative CGT, consistent with our result. While Gao et al. (2019) investigated the warming processes over these hotspots associated with the AMO circulation anomalies, they did not investigate the forcing mechanism of CGT by the AMO SST.

Our finding that the reanalysis SR is associated with North Atlantic SST anomalies that resemble the AMO pattern is also consistent with Lee et al. (2011). Lee et al. (2011) looked at the EOFs of the summertime upper-tropospheric circulation and the associated SSTs. Their EOF2 of the Northern Hemispheric 200-hPa geopotential height (their Fig. 3d) consists of a zonally symmetric component and a zonally asymmetric component resembling CGT. The positive phase of their observed EOF2 is associated with SST anomalies in the tropical Pacific, the North Pacific, and the North Atlantic (their Fig. 9b), similar to La Niña–like SSTs and North Atlantic SST anomalies that resemble the positive AMO pattern. They proposed that the zonally symmetric higher heights in the midlatitudes and zonally symmetric lower heights in the tropics (their Fig. 3d) is a response to the La Niña–like SSTs, consistent with Ding et al. (2011, their Fig. 8b). It is therefore likely that the zonally asymmetric negative CGT component in their EOF2 is related to North Atlantic SST anomalies that resemble the positive AMO pattern.

While this study, as well as those of Yasui and Watanabe (2010) and Kosaka et al. (2012), is able to find the relationship of SR with North Atlantic SST in reanalysis, Stephan et al. (2019) demonstrated that such relationship also exists in their simulation. Using an atmospheric model coupled to a global ocean mixed layer model, they allow for air–sea coupling with minimal SST bias relative to their desired mean ocean state. By constraining the mean ocean state in the North Atlantic to be relatively warmer or colder, they found that a colder North Atlantic slightly alters the SR phase distribution to become more positive, and a warmer North Atlantic slightly alters the distribution to become more negative (their Fig. 7b).

Another finding of our study is the association of ENSO-like SST with SR in the hindcast runs. Using a fully coupled atmosphere–ocean model, Stephan et al. (2019) also found SR to be associated with ENSO-like SST (their Figs. 6b–e). These ENSO-like SST are also correlated with ISM (their Figs. 11a–d), which has been proposed as a forcing for SR. However, our hindcast runs generally capture well the ENSO–ISM relationship compared to reanalysis. Therefore, the SR–ENSO relationship in our hindcast runs requires an explanation other than ISM.

We propose that basic-state bias in the hindcast runs can explain the SR–ENSO relationship. Analysis of the upper-level wind reveals biases in the stationary Rossby wave waveguide. Wind bias in the North Atlantic results in a discontinuous waveguide connecting North Atlantic and Asia. While we have not investigated how the North Atlantic SST can trigger waves, Yasui and Watanabe (2010) used reanalysis to show that colder North Atlantic SST (their Fig. 14) is associated with anomalous diabatic heating over North America (their Fig. 13a), which then forces a positive SR in their linear baroclinic model (their Fig. 13c). It is therefore likely that the discontinuous waveguide in our hindcast runs hinders wave propagation from North America into Asia.

Furthermore in our analysis, wind bias near the subtropical western Pacific may favor westward propagation of zonally elongated waves associated with ENSO into Asia. The idea that stationary Rossby waves can propagate westward from the Pacific in the summer has also been reported in other studies using barotropic model experiments, although not in an SR setting. Hoskins and Ambrizzi (1993, their Fig. 11) found that when a vorticity source is placed just east of the strong meridional gradient of absolute vorticity associated with the Asian jet, strong and rapid westward movement of vorticity of one sign is simulated in addition to the eastward wave development. While investigating the Indian monsoon and summertime ENSO, Shaman and Tziperman (2007) found circulation anomalies in the North African–Asian jet and attributed them to westward-propagating Rossby waves.

Our hindcast members that show the SR–ENSO relationship also show meridional wind shortwave anomalies linking the equatorial Pacific and the midlatitudes in Asia (members 1, 3, and 4 from Fig. 5). A question remains as to how to reconcile these short zonal waves with the westward-propagation theory, which allows only long zonal waves. To address this, we refer to the barotropic experiments by O’Reilly et al. (2018). They used a barotropic model to investigate the dominant covarying mode between summertime Euro-Atlantic circulation and circumglobal tropical precipitation. They analyzed the response to a forcing in the tropical eastern Pacific (their Fig. 6e), which is similar to our ENSO-like SST location. They found clear evidence of westward-propagating long zonal waves from the tropical eastern Pacific to the midlatitudes in Asia (their Fig. 11b). Short waves then develop over the west Pacific following the arrival of the long waves (their Fig. 11c). It is therefore likely that in our hindcast runs, the short waves in the west Pacific may be the response to the arrival of the westward-propagating long waves triggered by ENSO-like SST.

While all our ensemble members exhibit similar climatological upper-level wind bias (Fig. S7), as well as similar ENSO SST variability (Fig. 8), only weak intermember correlations in SR phases are found (Fig. S2). This is in line with the potential predictability of SR in our hindcast runs being weak (Fig. 4). This suggests that while ENSO modulates SR in our hindcast runs through upper-level wind bias, internal atmospheric dynamics still dominate over the oceanic modulation in generating SR. This agrees with Stephan et al. (2019, their Fig. 7b), who also only found a weak modulation of the SR phase distribution by SST.

In addition to the Asian jet affecting SR, SR can also affect the Asian jet. Hong and Lu (2016) identified a northward shift in the Asian jet during positive SR. This is because there is an unequal number of cyclonic and anticyclonic anomalies populated along the Asian jet. During positive SR, the two anticyclonic circulation anomalies over western and eastern Asia contribute to westerly anomalies north of the jet and easterly anomalies south of the jet. These outweigh the opposite contribution from the single cyclonic anomaly over central Asia, resulting in an overall northward jet shift. We find that these zonal wind anomalies associated with SR have an amplitude similar to the meridional wind anomalies (not shown), in agreement with the interpretation that the jet shift is just part of the SR circulation. Therefore, instead of including the effect of SR on the Asian jet, we only consider the climatological jet in our waveguide analysis.

Beverley et al. (2019) proposed several aspects of model errors that may be important for the generation of CGT, but the details are different from this study. First, whereas the Asian jet in their ECMWF hindcasts exhibits an obvious northward displacement (their Fig. 10), our hindcast runs show little latitudinal displacement (Fig. 9a). Therefore, the details of how the jet bias affects SR are expected to be different between our study and theirs. Second, they found that the midlatitude Rossby wave source in west-central Asia is biased to the north and also broader in space. Although Enomoto et al. (2003) have also suggested this RWS to be an important forcing region for SR, we are inclined to believe that this RWS is more of the result of the CGT bias than a cause, given its close proximity to the CGT centers of action. It seems likely that the northward shift in the jet resulted in a northward shift in CGT, thus shifting the RWS northward.

By comparing seasonal hindcast experiments of our coupled climate model with reanalysis, we have inferred the role of SST in triggering SR under the framework of stationary Rossby wave dispersions. We have concluded that the role of SST on SR can be substantially changed depending on the fidelity of model upper-level background winds. While we have proposed how the waveguides affect the dispersions, we have yet to discuss the role of the Rossby wave sources. It is our intention for a follow-up study to investigate more quantitatively the roles of the RWS and the waveguides, using idealized barotropic model experiments.

Acknowledgments

We thank the three anonymous reviewers whose comments have improved this manuscript. We acknowledge Dawson (2016) for the EOF codes. We acknowledge the GPCP Precipitation data provided by the NOAA/OAR/ESRL PSL, Boulder, Colorado, USA, from their Web site at https://psl.noaa.gov/. The appointment and study of Ronald Li are supported by The Chinese University of Hong Kong Impact Postdoctoral Fellowship Scheme and the Vice-Chancellor’s Discretionary Fund (Grants 3133132, 4930744). The appointment of Ngar-Cheung Lau at The Chinese University of Hong Kong is partially supported by the AXA Research Fund.

Data availability statement

Data underlying the findings can be accessed by contacting the corresponding author.

REFERENCES

  • Adler, R. F., and Coauthors, 2003: The version-2 Global Precipitation Climatology Project (GPCP) monthly precipitation analysis (1979–present). J. Hydrometeor., 4, 11471167, https://doi.org/10.1175/1525-7541(2003)004<1147:TVGPCP>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Ambrizzi, T., B. J. Hoskins, and H.-H. Hsu, 1995: Rossby wave propagation and teleconnection patterns in the austral winter. J. Atmos. Sci., 52, 36613672, https://doi.org/10.1175/1520-0469(1995)052<3661:RWPATP>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Bayasgalan, G., and J.-B. Ahn, 2018: Seasonal prediction of high-resolution temperature at 2-m height over Mongolia during boreal winter using both coupled general circulation model and artificial neural network. Int. J. Climatol., 38, 54185429, https://doi.org/10.1002/joc.5848.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Beverley, J. D., S. J. Woolnough, L. H. Baker, S. J. Johnson, and A. Weisheimer, 2019: The Northern Hemisphere circumglobal teleconnection in a seasonal forecast model and its relationship to European summer forecast skill. Climate Dyn., 52, 37593771, https://doi.org/10.1007/S00382-018-4371-4.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Dawson, A., 2016: eofs: A library for EOF analysis of meteorological, oceanographic, and climate data. J. Open Res. Software, 4, p.e14, https://doi.org/10.5334/JORS.122.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Ding, Q., and B. Wang, 2005: Circumglobal teleconnection in the Northern Hemisphere summer. J. Climate, 18, 34833505, https://doi.org/10.1175/JCLI3473.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Ding, Q., B. Wang, J. M. Wallace, and G. Branstator, 2011: Tropical–extratropical teleconnections in boreal summer: Observed interannual variability. J. Climate, 24, 18781896, https://doi.org/10.1175/2011JCLI3621.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Enomoto, T., B. J. Hoskins, and Y. Matsuda, 2003: The formation mechanism of the Bonin high in August. Quart. J. Roy. Meteor. Soc., 129, 157178, https://doi.org/10.1256/qj.01.211.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Gao, M., J. Yang, D. Gong, P. Shi, Z. Han, and S.-J. Kim, 2019: Footprints of Atlantic multidecadal oscillation in the low-frequency variation of extreme high temperature in the Northern Hemisphere. J. Climate, 32, 791802, https://doi.org/10.1175/JCLI-D-18-0446.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Hersbach, H., and Coauthors, 2019: Global reanalysis: Goodbye ERA-Interim, hello ERA5. ECMWF Newsletter, No. 159, ECMWF, Reading, United Kingdom, 17–24, https://www.ecmwf.int/node/19027.

  • Hollingsworth, A., K. Arpe, M. Tiedtke, M. Capaldo, and H. Savijärvi, 1980: The performance of a medium-range forecast model in winter—Impact of physical parameterizations. Mon. Wea. Rev., 108, 17361773, https://doi.org/10.1175/1520-0493(1980)108<1736:TPOAMR>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Hong, X., and R. Lu, 2016: The meridional displacement of the summer Asian jet, Silk Road pattern, and tropical SST anomalies. J. Climate, 29, 37533766, https://doi.org/10.1175/JCLI-D-15-0541.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Hoskins, B. J., and D. J. Karoly, 1981: The steady linear response of a spherical atmosphere to thermal and orographic forcing. J. Atmos. Sci., 38, 11791196, https://doi.org/10.1175/1520-0469(1981)038<1179:TSLROA>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Hoskins, B. J., and T. Ambrizzi, 1993: Rossby wave propagation on a realistic longitudinally varying flow. J. Atmos. Sci., 50, 16611671, https://doi.org/10.1175/1520-0469(1993)050<1661:RWPOAR>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kim, H.-J., and J.-B. Ahn, 2015: Improvement in prediction of the Arctic Oscillation with a realistic ocean initial condition in a CGCM. J. Climate, 28, 89518967, https://doi.org/10.1175/JCLI-D-14-00457.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kosaka, Y., H. Nakamura, M. Watanabe, and M. Kimoto, 2009: Analysis on the dynamics of a wave-like teleconnection pattern along the summertime Asian jet based on a reanalysis dataset and climate model simulations. J. Meteor. Soc. Japan, 87, 561580, https://doi.org/10.2151/jmsj.87.561.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kosaka, Y., J. S. Chowdary, S.-P. Xie, Y.-M. Min, and J.-Y. Lee, 2012: Limitations of seasonal predictability for summer climate over East Asia and the northwestern Pacific. J. Climate, 25, 75747589, https://doi.org/10.1175/JCLI-D-12-00009.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Lee, J.-Y., B. Wang, Q. Ding, K.-J. Ha, J.-B. Ahn, A. Kumar, B. Stern, and O. Alves, 2011: How predictable is the Northern Hemisphere summer upper-tropospheric circulation? Climate Dyn., 37, 11891203, https://doi.org/10.1007/S00382-010-0909-9.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Li, R. K. K., T. Woollings, C. O’Reilly, and A. A. Scaife, 2020: Effect of the North Pacific tropospheric waveguide on the fidelity of model El Niño teleconnections. J. Climate, 33, 52235237, https://doi.org/10.1175/JCLI-D-19-0156.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • North, G. R., T. L. Bell, R. F. Cahalan, and F. J. Moeng, 1982: Sampling errors in the estimation of empirical orthogonal functions. Mon. Wea. Rev., 110, 699706, https://doi.org/10.1175/1520-0493(1982)110<0699:SEIEO>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • O’Reilly, C. H., T. Woollings, L. Zanna, and A. Weisheimer, 2018: The impact of tropical precipitation on summertime Euro-Atlantic circulation via a circumglobal wave train. J. Climate, 31, 64816504, https://doi.org/10.1175/JCLI-D-17-0451.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Rowell, D. P., C. K. Folland, K. Maskell, and M. N. Ward, 1995: Variability of summer rainfall over tropical North Africa (1906–92): Observations and modelling. Quart. J. Roy. Meteor. Soc., 121, 669704, https://doi.org/10.1002/QJ.49712152311.

    • Search Google Scholar
    • Export Citation
  • Shaman, J., and E. Tziperman, 2007: Summertime ENSO–North African–Asian jet teleconnection and implications for the Indian monsoons. Geophys. Res. Lett., 34, L11702, https://doi.org/10.1029/2006GL029143.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Stephan, C. C., N. P. Klingaman, and A. G. Turner, 2019: A mechanism for the recently increased interdecadal variability of the Silk Road pattern. J. Climate, 32, 717736, https://doi.org/10.1175/JCLI-D-18-0405.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Sun, J., and J.-B. Ahn, 2015: Dynamical seasonal predictability of the Arctic Oscillation using a CGCM. Int. J. Climatol., 35, 13421353, https://doi.org/10.1002/joc.4060.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Webster, P. J., V. O. Magaña, T. N. Palmer, J. Shukla, R. A. Tomas, M. Yanai, and T. Yasunari, 1998: Monsoons: Processes, predictability, and the prospects for prediction. J. Geophys. Res., 103, 14 45114 510, https://doi.org/10.1029/97JC02719.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Yasui, S., and M. Watanabe, 2010: Forcing processes of the summertime circumglobal teleconnection pattern in a dry AGCM. J. Climate, 23, 20932114, https://doi.org/10.1175/2009JCLI3323.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Zhou, F., R. Zhang, and J. Han, 2019: Relationship between the circumglobal teleconnection and Silk Road pattern over Eurasian continent. Sci. Bull., 64, 374376, https://doi.org/10.1016/j.scib.2019.02.014.

    • Crossref
    • Search Google Scholar
    • Export Citation
Save