• Allen, D. R., , L. Coy, , S. D. Eckermann, , J. P. McCormack, , G. L. Manney, , T. F. Hogan, , and Y.-J. Kim, 2006: NOGAPS-ALPHA simulations of the 2002 Southern Hemisphere stratospheric major warming. Mon. Wea. Rev., 134, 498518, doi:10.1175/MWR3086.1.

    • Search Google Scholar
    • Export Citation
  • Andrews, D. G., , J. R. Holton, , and C. B. Leovy, 1987: Middle Atmosphere Dynamics. International Geophysics Series, Vol. 40, Academic Press, 489 pp.

  • Baldwin, M. P., , and T. J. Dunkerton, 1999: Propagation of the Arctic Oscillation from the stratosphere to the troposphere. J. Geophys. Res., 104, 30 93730 946, doi:10.1029/1999JD900445.

    • Search Google Scholar
    • Export Citation
  • Baldwin, M. P., , and T. J. Dunkerton, 2001: Stratospheric harbingers of anomalous weather regimes. Science, 294, 581584, doi:10.1126/science.1063315.

    • Search Google Scholar
    • Export Citation
  • Baldwin, M. P., , and D. W. J. Thompson, 2009: A critical comparison of stratosphere–troposphere coupling indices. Quart. J. Roy. Meteor. Soc., 135, 16611672, doi:10.1002/qj.479.

    • Search Google Scholar
    • Export Citation
  • Baldwin, M. P., , T. Hirooka, , A. O’Neill, , S. Yoden, , A. J. Charlton, , Y. Hio, , W. A. Lahoz, , and A. Mori, 2003a: Major stratospheric warming in the Southern Hemisphere in 2002: Dynamical aspects of the ozone hole split. SPARC Newsletter, No. 20, SPARC Office, Verrières-le-Buisson, France, 2226.

  • Baldwin, M. P., , D. B. Stephenson, , D. W. J. Thompson, , T. J. Dunkerton, , A. J. Charlton, , and A. O’Neill, 2003b: Stratospheric memory and extended-range weather forecasts. Science, 301, 636640, doi:10.1126/science.1087143.

    • Search Google Scholar
    • Export Citation
  • Charlton, A. J., , and L. M. Polvani, 2007: A new look at stratospheric sudden warmings. Part I. Climatology and modeling benchmarks. J. Climate, 20, 449469, doi:10.1175/JCLI3996.1.

    • Search Google Scholar
    • Export Citation
  • Charlton, A. J., , and D. Jackson, 2012: SNAP: The Stratospheric Network for the Assessment of Predictability. SPARC Newsletter, No. 39, SPARC Office, Zurich, Switzerland, 4042.

  • Chen, P., , and W. A. Robinson, 1992: Propagation of planetary waves between the troposphere and stratosphere. J. Atmos. Sci., 49, 25332545, doi:10.1175/1520-0469(1992)049<2533:POPWBT>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Harada, Y., , A. Goto, , H. Hasegawa, , N. Fujikawa, , H. Naoe, , and T. Hirooka, 2010: A major stratospheric sudden warming event in January 2009. J. Atmos. Sci., 67, 20522069, doi:10.1175/2009JAS3320.1.

    • Search Google Scholar
    • Export Citation
  • Hirooka, T., , T. Ichimaru, , and H. Mukougawa, 2007: Predictability of stratospheric sudden warmings as inferred from ensemble forecast data: Intercomparison of 2001/02 and 2003/04 winters. J. Meteor. Soc. Japan, 85, 919925, doi:10.2151/jmsj.85.919.

    • Search Google Scholar
    • Export Citation
  • Hirota, I., , and Y. Sato, 1969: Periodic variation of the winter stratospheric circulation and intermittent vertical propagation of planetary waves. J. Meteor. Soc. Japan, 47, 390402.

    • Search Google Scholar
    • Export Citation
  • Iwasaki, T., , S. Yamada, , and K. Tada, 1989: A parameterization scheme of orographic gravity wave drag with two different vertical partitionings. Part I: Impacts on medium-range forecasts. J. Meteor. Soc. Japan, 67, 1127.

    • Search Google Scholar
    • Export Citation
  • Kim, Y.-J., , and M. Flatau, 2010: Hindcasting the January 2009 Arctic sudden stratospheric warming and its influence on the Arctic Oscillation with unified parameterization of orographic drag in NOGAPS. Part I: Extended-range stand-alone forecast. Wea. Forecasting, 25, 16281644, doi:10.1175/2010WAF2222421.1.

    • Search Google Scholar
    • Export Citation
  • Labitzke, K. G., , and H. van Loon, 1999: The Stratosphere: Phenomena, History, and Relevance. Springer, 179 pp.

  • Lee, J. N., , D. L. Wu, , G. L. Manney, , and M. J. Schwartz, 2009: Aura Microwave Limb Sounder observations of the northern annular mode: From the mesosphere to the upper troposphere. Geophys. Res. Lett., 36, L20807, doi:10.1029/2009GL040678.

    • Search Google Scholar
    • Export Citation
  • Limpasuvan, V., , D. W. J. Thompson, , and D. L. Hartmann, 2004: The life cycle of Northern Hemisphere sudden stratospheric warmings. J. Climate, 17, 25842596, doi:10.1175/1520-0442(2004)017<2584:TLCOTN>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Manney, G. L., and Coauthors, 2005: Simulations of dynamics and transport during the September 2002 Antarctic major warming. J. Atmos. Sci., 62, 690707, doi:10.1175/JAS-3313.1.

    • Search Google Scholar
    • Export Citation
  • Manney, G. L., and Coauthors, 2009: Aura Microwave Limb Sounder observations of dynamics and transport during the record-breaking 2009 Arctic stratospheric major warming. Geophys. Res. Lett., 36, L12815, doi:10.1029/2009GL038586.

    • Search Google Scholar
    • Export Citation
  • Marshall, A. G., , and A. A. Scaife, 2010: Improved predictability of stratospheric sudden warming events in an atmospheric general circulation model with enhanced stratospheric resolution. J. Geophys. Res., 115, D16114, doi:10.1029/2009JD012643.

    • Search Google Scholar
    • Export Citation
  • Matsuno, T., 1971: A dynamical model of the stratospheric sudden warming. J. Atmos. Sci., 28, 14791494, doi:10.1175/1520-0469(1971)028<1479:ADMOTS>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Mukougawa, H., , and T. Hirooka, 2004: Predictability of stratospheric sudden warming: A case study for 1998/99 winter. Mon. Wea. Rev., 132, 17641776, doi:10.1175/1520-0493(2004)132<1764:POSSWA>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Mukougawa, H., , H. Sakai, , and T. Hirooka, 2005: High sensitivity to the initial condition for the prediction of stratospheric sudden warming. Geophys. Res. Lett., 32, L17806, doi:10.1029/2005GL022909.

    • Search Google Scholar
    • Export Citation
  • Mukougawa, H., , T. Hirooka, , and Y. Kuroda, 2009: Influence of stratospheric circulation on the predictability of the tropospheric northern annular mode. Geophys. Res. Lett., 36, L08814, doi:10.1029/2008GL037127.

    • Search Google Scholar
    • Export Citation
  • Nakagawa, K. I., , and K. Yamazaki, 2006: What kind of stratospheric sudden warming propagates to the troposphere? Geophys. Res. Lett., 33, L04801, doi:10.1029/2005GL024784.

    • Search Google Scholar
    • Export Citation
  • Nishii, K., , and H. Nakamura, 2004: Tropospheric influence on the diminished Antarctic ozone hole in September 2002. Geophys. Res. Lett., 31, L16103, doi:10.1029/2004GL019532.

    • Search Google Scholar
    • Export Citation
  • O’Neill, A., 2003: Stratospheric sudden warmings. Encyclopedia of Atmospheric Sciences, J. R. Holton, J. A. Curry, and J. A. Pyle, Eds., Vol. 4, Academic Press, 1342–1353.

  • Onogi, K., and Coauthors, 2007: The JRA-25 Reanalysis. J. Meteor. Soc. Japan, 85, 369432, doi:10.2151/jmsj.85.369.

  • Plumb, R. A., 1985: On the three-dimensional propagation of stationary waves. J. Atmos. Sci., 42, 217229, doi:10.1175/1520-0469(1985)042<0217:OTTDPO>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Polvani, L. M., , and D. W. Waugh, 2004: Upward wave activity flux as a precursor to extreme stratospheric events and subsequent anomalous surface weather regimes. J. Climate, 17, 35483554, doi:10.1175/1520-0442(2004)017<3548:UWAFAA>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Shiotani, M., 1986: Planetary wave activity in the troposphere and stratosphere during the Northern Hemisphere winter. J. Atmos. Sci., 43, 32003209, doi:10.1175/1520-0469(1986)043<3200:PWAITT>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Sigmond, M., , and J. F. Scinocca, 2010: The influence of basic state on the Northern Hemisphere circulation response to climate change. J. Climate, 23, 14341446, doi:10.1175/2009JCLI3167.1.

    • Search Google Scholar
    • Export Citation
  • Sigmond, M., , J. F. Scinocca, , V. V. Kharin, , and T. G. Shepherd, 2013: Enhanced seasonal forecast skill following stratospheric sudden warmings. Nat. Geosci., 6, 98102, doi:10.1038/ngeo1698.

    • Search Google Scholar
    • Export Citation
  • Simmons, A., , M. Hortal, , G. Kelly, , A. McNally, , A. Untch, , and S. Uppala, 2005: ECMWF analyses and forecasts of stratospheric winter polar vortex breakup: September 2002 in the Southern Hemisphere and related events. J. Atmos. Sci., 62, 668689, doi:10.1175/JAS-3322.1.

    • Search Google Scholar
    • Export Citation
  • Stan, C., , and D. M. Straus, 2009: Stratospheric predictability and sudden stratospheric warming events. J. Geophys. Res., 114, D12103, doi:10.1029/2008JD011277.

    • Search Google Scholar
    • Export Citation
  • Sun, L., , W. A. Robinson, , and G. Chen, 2012: The predictability of stratospheric warming events: More from the troposphere or the stratosphere? J. Atmos. Sci., 69, 768783, doi:10.1175/JAS-D-11-0144.1.

    • Search Google Scholar
    • Export Citation
  • Taguchi, M., 2014: Stratospheric predictability: Basic characteristics in JMA 1-month hindcast experiments for 1979–2009. J. Atmos. Sci., doi:10.1175/JAS-D-13-0295.1, in press.

    • Search Google Scholar
    • Export Citation
  • Thompson, D. W. J., , and J. M. Wallace, 1998: The Arctic Oscillation signature in the wintertime geopotential height and temperature fields. Geophys. Res. Lett., 25, 12971300, doi:10.1029/98GL00950.

    • Search Google Scholar
    • Export Citation
  • Thompson, D. W. J., , and J. M. Wallace, 2000: Annular modes in the extratropical circulation. Part I: Month-to-month variability. J. Climate, 13, 10001016, doi:10.1175/1520-0442(2000)013<1000:AMITEC>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Tibaldi, S., , and F. Molteni, 1990: On the operational predictability of blocking. Tellus, 42A, 343365, doi:10.1034/j.1600-0870.1990.t01-2-00003.x.

    • Search Google Scholar
    • Export Citation
  • Woollings, T., , A. Charlton-Perez, , S. Ineson, , A. G. Marshall, , and G. Masato, 2010: Associations between stratospheric variability and tropospheric blocking. J. Geophys. Res., 115, D06108, doi:10.1029/2009JD012742.

    • Search Google Scholar
    • Export Citation
  • Yoden, S., , T. Yamaga, , S. Pawson, , and U. Langematz, 1999: A composite analysis of the stratospheric sudden warmings simulated in a perpetual January integration of the Berlin TSM GCM. J. Meteor. Soc. Japan, 77, 431445.

    • Search Google Scholar
    • Export Citation
  • View in gallery

    Zonal mean zonal wind at 60°N/S, 10 hPa for the three MSSWs as a function of time lag: 2002 (solid), 2009 (dash), and 1989 (dashed–dotted). Lag-0 day corresponds to the key days when the easterly winds maximize. Black ticks denote the initial dates of the HC experiments examined in this study (see Table 1).

  • View in gallery

    (a)–(c) Zonal mean zonal wind at 60°N/S, 10 hPa and (d)–(f) poleward eddy heat flux in 45°–75°N/S, 100 hPa for the reanalysis and HC data as functions of time lag. Black lines denote the reanalysis data. Color lines denote the HC data: green for the HC A group, blue for B group, and purple for C group. Thick lines denote the ensemble means, and thin lines denote respective ensemble members. Cases shown are (a),(d) 2002; (b),(e) 2009; and (c),(f) 1989.

  • View in gallery

    Scatterplots of cumulative poleward eddy heat flux in 45°–75°N/S, 100 hPa (x axis) and mean zonal wind at 60°N/S, 10 hPa (y axis). The heat flux is cumulative from lag = −20 to 0 days. The zonal wind is averaged between lag = −2 and +2 days. Black squares denote the reanalysis data. Color markers denote the HC data: green for HC A group, blue for B group, and purple for C group. Larger (filled) markers denote the ensemble means, and smaller (open) markers denote the respective ensemble members. Horizontal and vertical solid lines denote 95% confidence intervals of each quantity based on the Student’s t distribution for the five members. The reanalysis data are substituted before the initial dates when HC data are needed but unavailable in cumulating the heat flux.

  • View in gallery

    Longitude–height sections of the WAF (arrows and gray shades) of quasi-stationary waves 1–3 in the (a),(c),(e) reanalysis data and (b),(d),(f) HC B data. The zonal and vertical components of the WAF for 45°–75°N/S for lag = −10 to 0 days are plotted. The WAF of the HC data is calculated from the ensemble-mean fields. Light and dark gray shades denote where the vertical component exceeds 0.04 and 0.08 m2 s−2, respectively. The quasi-stationary waves are extracted with a 5-day running mean. The geopotential height waves 1–3 are also shown by black contours for the same latitudinal and time-lag conditions. Cases shown are (a),(b) 2002; (c),(d) 2009; and (e),(f) 1989.

  • View in gallery

    Time lag–longitude sections of the quasi-stationary geopotential height waves 1–3 for 45°–75°N/S at 300 hPa. Shown are (a)–(c) the reanalysis data and (d)–(f) the ensemble means of the HC B data for cases (a),(d) 2002; (b),(e) 2009; and (c),(f) 1989. Contour interval is 100 m. Gray and white dots denote where the WAFz at 100 hPa exceeds 0.04 and 0.08 m2 s−2, respectively.

  • View in gallery

    Synoptic maps of the low-passed geopotential height at 300 hPa for lag = −5 ± 2 days in (a)–(c) the reanalysis data and (d)–(f) the ensemble means of the HC B data. Black contours denote the total fields (zonal mean and all waves; contour interval is 200 m), and color shades denote waves 1–3 (contour interval is 100 m; warm and cold colors denote positive and negative values, respectively). Shown are cases (a),(d) 2002; (b),(e) 2009; and (c),(f) 1989.

  • View in gallery

    Latitude–height sections of the mean zonal wind (thick contours; m s−1) and refractive index for wave 2 (thin contours and shaded) for lag = −5 ± 2 days of the 2009 case for (a) the reanalysis data and (b) the ensemble mean of the HC B data. Light and dark shades denote where the refractive index (multiplied by the square of Earth’s radius) exceeds 80 and 160, respectively. (c) The zonal wind difference [(b) minus (a)]; contours are drawn with a 2 m s−1 interval for negative values and a 10 m s−1 interval for zero and positive values, with additional contours at +5 m s−1. (d),(e) As in (a),(b), but for zonal means of the meridional and vertical components of WAF. Light and dark shades denote where the zonal mean WAFz exceed 0.04 and 0.08 m2 s−2, respectively.

  • View in gallery

    Time lag–latitude sections of the mean zonal wind (gray contours and shaded) and Eliassen–Palm flux divergence (black contours drawn at −12, −8, and −4 m s−1 day−1) for the (a),(c) 2009 and (b),(d) 1989 MSSWs in (a),(b) the reanalysis data and (c),(d) the ensemble means of the HC B data. Both quantities are averaged over 200–500 hPa. Light shades are applied to wind values between 0 and 10 m s−1 and dark shades are applied to those above 10 m s−1. The Eliassen–Palm flux divergence is smoothed in time and latitude for graphical purposes.

  • View in gallery

    As in Fig. 2, but for the zonal mean geopotential height (m) averaged poleward of 65°N/S (a)–(c) at 10 hPa and (d)–(f) near the surface. Heights are (d) 850 hPa (1100 m is subtracted for graphical purposes) and (e),(f) 1000 hPa. Results from the HC D group are added in red.

  • View in gallery

    As in Fig. 3, but for the normalized zonal mean geopotential height anomalies poleward of 65°N/S. The x axis uses the 10-hPa height for (a)–(c) lag = −2 to +2 days and (d)–(f) lag = +1 to +15 days. The y axis uses near-surface height for lag = +1 to +15 days: (a),(d) 850 and (b),(c),(e),(f) 1000 hPa. The results are colored as in Fig. 9. The HC A data are excluded, owing to insufficient data after the MSSWs. The reanalysis data are substituted before the initial dates, when HC data are needed but unavailable in calculating the time means, as in Fig. 3. The same treatment is also applied in Figs. 12 and 13.

  • View in gallery

    Time–height sections of the normalized zonal mean geopotential height anomalies averaged poleward of 65°N/S for the three cases: (a),(d),(g),(j) 2002; (b),(e),(h),(k) 2009; and (c),(f),(i),(l) 1989 for (a)–(c) the reanalysis data and the ensemble means of (d)–(f) HC B, (g)–(i) HC C, and (j)–(l) HC D. Contour interval is 0.5. Gray dots in (d)–(l) indicate that the differences of the HC data from the JRA-25 data are statistically significant at the 95% confidence level (see the text for details). Gray dots are marked only where the differences of the HC ensemble mean from the JRA-25 data exceed ±0.5.

  • View in gallery

    Synoptic maps of the near-surface geopotential height for lag = +1 to +15 days of the three MSSWs: (a),(d),(g),(j) 2002 case at 850 hPa, (b),(e),(h),(k) 2009 at 1000 hPa, and (c),(f),(i),(l) 1989 at 1000 hPa for (a)–(c) the reanalysis data and the ensemble means of (d)–(f) HC B, (g)–(i) HC C, and (j)–(l) HC D. Colors denote anomalies from the climatology of the reanalysis data (contour interval is 50 m). Black contours denote differences of the HC data from the reanalysis data (contour interval is 50 m). Coastal lines (white) and reference longitudes are drawn in (a)–(c).

  • View in gallery

    Vertical profiles of anomaly correlations between the reanalysis and HC B–D data for the three MSSWs: (a) 2002, (b) 2009, and (c) 1989. The results are colored as in Fig. 9. The anomaly correlations are calculated for geopotential height anomalies poleward of 20°N/S for lag = +1 to +15 days between the reanalysis and each ensemble member, and 95% confidence intervals are obtained with Student’s t distribution (denoted by horizontal lines at each level). The horizontal lines are slightly offset for graphical purposes. The markers (triangles, circles, and diamonds) denote the anomaly correlations of the ensemble mean fields.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 28 28 6
PDF Downloads 20 20 0

Predictability of Major Stratospheric Sudden Warmings of the Vortex Split Type: Case Study of the 2002 Southern Event and the 2009 and 1989 Northern Events

View More View Less
  • 1 Department of Earth Science, Aichi University of Education, Kariya, Japan
© Get Permissions
Full access

Abstract

This study investigates the predictability of three major stratospheric sudden warmings (MSSWs) of the vortex split type: the Southern Hemisphere case in September 2002 and two Northern Hemisphere cases in January 2009 and February 1989. The author examines changes in the predictability of the MSSWs with lead time, as well as the connection of the predictability to lower-atmospheric features for pre- and post-MSSW periods. The Japan Meteorological Agency (JMA)’s 1-month ensemble hindcast (HC) experiment data are compared to the Japanese 25-year Reanalysis Project (JRA-25)/JMA Climate Data Assimilation System (JCDAS) data.

For the pre-MSSW period, a strong predictability connection is observed among all three cases. Unsuccessful predictions of the MSSWs are characterized by an underestimation (or lack) of the enhanced wave activity in the lower stratosphere, which is further related to the strength and persistence of the upper-tropospheric ridge and trough. The mean zonal wind profile in the upper troposphere is also important for the 2009 case. These results confirm the role of tropospheric wave forcing of the MSSWs in the context of predictability. The characteristic time scale for successful predictions is approximately 10 days–2 weeks, which roughly corresponds to the time scale of the tropospheric wave forcing. No ensemble member successfully predicts the MSSWs with lead times longer than the time scale.

The predictability connection between the stratospheric and tropospheric anomalies is more subtle for the post-MSSW period. In particular, the HC group initialized about 1 week before the MSSWs tends to reproduce the evolution of the stratosphere after the MSSWs well but not that of the troposphere in some cases.

Corresponding author address: Masakazu Taguchi, Department of Earth Science, Aichi University of Education, 1 Hirosawa, Igaya-cho, Kariya 448-8542, Japan. E-mail: mtaguchi@auecc.aichi-edu.ac.jp

Abstract

This study investigates the predictability of three major stratospheric sudden warmings (MSSWs) of the vortex split type: the Southern Hemisphere case in September 2002 and two Northern Hemisphere cases in January 2009 and February 1989. The author examines changes in the predictability of the MSSWs with lead time, as well as the connection of the predictability to lower-atmospheric features for pre- and post-MSSW periods. The Japan Meteorological Agency (JMA)’s 1-month ensemble hindcast (HC) experiment data are compared to the Japanese 25-year Reanalysis Project (JRA-25)/JMA Climate Data Assimilation System (JCDAS) data.

For the pre-MSSW period, a strong predictability connection is observed among all three cases. Unsuccessful predictions of the MSSWs are characterized by an underestimation (or lack) of the enhanced wave activity in the lower stratosphere, which is further related to the strength and persistence of the upper-tropospheric ridge and trough. The mean zonal wind profile in the upper troposphere is also important for the 2009 case. These results confirm the role of tropospheric wave forcing of the MSSWs in the context of predictability. The characteristic time scale for successful predictions is approximately 10 days–2 weeks, which roughly corresponds to the time scale of the tropospheric wave forcing. No ensemble member successfully predicts the MSSWs with lead times longer than the time scale.

The predictability connection between the stratospheric and tropospheric anomalies is more subtle for the post-MSSW period. In particular, the HC group initialized about 1 week before the MSSWs tends to reproduce the evolution of the stratosphere after the MSSWs well but not that of the troposphere in some cases.

Corresponding author address: Masakazu Taguchi, Department of Earth Science, Aichi University of Education, 1 Hirosawa, Igaya-cho, Kariya 448-8542, Japan. E-mail: mtaguchi@auecc.aichi-edu.ac.jp

1. Introduction

Stratospheric sudden warmings (SSWs) are a spectacular phenomenon in the winter stratosphere especially for the Northern Hemisphere, where the polar vortex largely deforms and sometimes even breaks down during some winters (Andrews et al. 1987; O’Neill 2003; Labitzke and van Loon 1999). The deformation and/or breakdown of the polar vortex accompany a rapid increase in polar temperatures and a weakening or reversal of the westerly polar night jet.

There are at least two classifications of SSWs: one is based on the strength of the polar vortex (including direction of the polar night jet) during SSWs and the other is based on the morphology of the vortex. In the first classification, SSWs are called major SSWs (MSSWs) when the zonal mean zonal wind at 60°N/S, 10 hPa becomes easterly during the events. The other SSWs, for which the meridional temperature gradient, not the zonal wind, reverses (increases poleward of 60°N/S at 10 hPa), are called minor SSWs (Andrews et al. 1987). In this paper the acronym “SSWs” is used when major and minor warmings are not distinguished. In the second classification, SSWs are divided into two groups: vortex displacement SSWs and vortex split SSWs (Charlton and Polvani 2007). In vortex displacement SSWs, the vortex is merely displaced off from the pole without a split. In vortex split SSWs, on the other hand, the polar vortex is split into two cyclonic vortices. Generally, a planetary wave of zonal wavenumber 1 (wave 1) is dominant in vortex displacement events, whereas wave 2 is dominant in vortex split events.

Extensive diagnostic studies have demonstrated that SSWs are a representative example of coupled variability between the extratropical stratosphere and troposphere. A series of composite studies on SSWs using observations and general circulation models (GCMs) show both upward and downward influences in association with SSWs (e.g., Limpasuvan et al. 2004; Yoden et al. 1999). The upward influence essentially emerges as an upward propagation of enhanced planetary wave activity from the troposphere to the stratosphere. This feature is consistent with the assumption in the dynamical model of Matsuno (1971) for SSWs. The downward influence emerges as a downward propagation of circulation anomalies from the stratosphere to the troposphere (Baldwin and Dunkerton 1999, 2001). In the troposphere, circulation anomalies tend to persist for months, which correspond to the negative phase of the northern annular mode (Thompson and Wallace 1998, 2000). Woollings et al. (2010) also show significant correlations with the zonal mean flow in the stratosphere leading the occurrence of blocking in high latitudes.

Recent studies also investigated the predictability of stratospheric variability, such as SSWs, from case study and statistical perspectives.

Mukougawa and Hirooka (2004) examined the predictability of the vortex displacement MSSW in December 1998 with the operational 1-month forecast data of the Japan Meteorological Agency (JMA). They claimed that the MSSW is predictable by the system from 1 month in advance when it captures the generation of an enhanced wave 1 in the troposphere. Using all ensemble members of the same system, Mukougawa et al. (2005) showed that the vortex displacement MSSW in December 2001 is predictable from at least 2 weeks in advance and observed high sensitivity of the predictability to initial conditions: the predictability changes with the zonal wind structure that developed differently over the first week of the forecasts among different ensemble members. Hirooka et al. (2007) compared two vortex displacement MSSWs in December 2001 and December 2003, and found a significant difference in the predictable period between the two cases (16 and 9 days, respectively). They attributed the difference in the predictable period to the time evolution of the MSSWs: the period is shorter when the evolution is more complicated.

Allen et al. (2006) examined hindcast (HC) data of GCMs for the September 2002 MSSW in the Southern Hemisphere. This MSSW was a strikingly exceptional case that occurred in the Southern Hemisphere for the first time on record (e.g., Baldwin et al. 2003a). Basically, SSWs are a phenomenon in the Northern Hemisphere, where almost all dramatic events have occurred. Extensive studies have documented dynamical features of the MSSW (see special issue of Journal of Atmospheric Sciences, 2005, Vol. 62, No. 3). Allen et al. (2006) demonstrated that the forecasts up to 6 days capture the main features of the MSSW, but those beyond 6 days degrade the reproducibility (see also Simmons et al. 2005). The longer unsuccessful forecasts of Allen et al. underestimated the upward propagation of the planetary wave activity emanating from the blocking pattern over the South Atlantic. This supports the observational argument by Nishii and Nakamura (2004) that the blocking event plays an important role in forcing the planetary waves, and hence MSSW. This argument is further supported by Manney et al. (2005), who noted the importance of wave forcing at 100 hPa in reproducing the MSSW with a mechanistic model. Allen et al. (2006) also claimed that the forecasting skill during September 2002 is improved by a higher model top, better physical parameterizations, and better initial conditions.

Kim and Flatau (2010) performed HC experiments for the 2009 MSSW with the Navy Operational Global Atmospheric Prediction System (NOGAPS) model, where orographic drag parameterizations were changed. This MSSW was the strongest and most prolonged case on record (Manney et al. 2009). The predictability of the MSSW was improved up to a lead time of approximately 2 weeks by improving the orographic drag scheme. The improvement includes increases in the gravity wave drag at upper levels (stratosphere and mesosphere) at the expense of decreases in blocked-layer drag near the surface resulting from the blocking of horizontal flow by subgrid-scale orography. The reproducibility was associated with the enhanced heat flux in the lower stratosphere, affected by subgrid-scale orographic drag processes. Harada et al. (2010) observed that the enhanced heat flux is induced by a tropospheric ridge over Alaska. The successful prediction of the MSSW was followed by downward propagation and tropospheric circulation anomalies that resemble the observations (Lee et al. 2009; Harada et al. 2010).

Using the archive of the 1-month ensemble forecast of JMA from 2001/02 to 2005/06, the statistical study of T. Ichimaru et al. (2013, personal communication) revealed the nature of stratospheric predictability: the average predictable period in the stratosphere is longer than that in the troposphere; the period in the stratosphere shows large variability depending on dynamical conditions (shorter when the stratosphere is more dynamically active). Stan and Straus (2009) conducted HC GCM experiments for 10 Northern Hemisphere winters from 1981 to 1990. They pointed out that errors in the phase of planetary waves importantly affect the predictable period of stratospheric circulation, including SSWs. Marshall and Scaife (2010) reported maximum lead times, ranging from 9 to 15 days, of successful predictions for four MSSWs using their 60-level version of the Hadley Centre Global Environmental Model.

Stratospheric predictability has also drawn significant attention, as the knowledge of stratospheric conditions, such as SSW-associated anomalies, may be useful for improving extended-range weather forecasts (e.g., Baldwin and Dunkerton 2001; Baldwin et al. 2003b). In particular, Sigmond et al. (2013) demonstrated enhanced forecast skill for tropospheric circulation patterns in HC experiments with a dynamical forecast system. Sigmond et al. compared forecasts initialized at the onset date of SSWs with those initialized apart from SSWs. Using the 5-yr archive of the JMA 1-month ensemble forecast, Mukougawa et al. (2009) found that the prediction skill of the northern annular mode (NAM) index in the upper troposphere is improved for about a 10-day forecast when largely negative NAM indices are observed around 30 hPa at the initial time of forecast. Charlton and Jackson (2012) proposed a coordinated activity [Stratospheric Network for the Assessment of Predictability (SNAP)] to examine stratospheric predictability and its tropospheric impact in multiple systems.

This paper presents results from a case study that aims to investigate predictability features of three extreme MSSWs of the vortex split type: the September 2002 case in the Southern Hemisphere and the February 1989 and January 2009 cases in the Northern Hemisphere. To be more specific, we seek to examine two aspects of the predictability of the three MSSWs: what is the characteristic time scale for successful predictions of the MSSWs (in other words, how far in advance can the MSSWs be predicted for the pre-MSSW period, or more generally, how does the predictability of the MSSWs change with lead time?), and how is the predictability of the MSSWs connected to lower-atmospheric features (in the lower stratosphere and troposphere) for the pre- and post-MSSW periods? This study clarifies more concrete, dynamical features of the MSSW predictability through examining these questions. To this end, we mainly compare the JMA 1-month HC experiment data with the Japanese 25-year Reanalysis Project (JRA-25)/JMA Climate Data Assimilation System (JCDAS) data. By “predictability” we mean how well the HC data reproduce observed features in the real world, or the reanalysis data. Whereas this study is a pilot case study of the relatively recent, extreme MSSWs using the HC data, a statistical study with more samples is a natural extension; we are in fact conducting such an attempt, and will report the results in a separate follow-up paper (Taguchi 2014).

The motivation for this study is that these aspects of the stratospheric predictability need to be further investigated for a few extreme cases. The three cases are therefore chosen here, because they are extreme but relatively unexplored in terms of predictability. The 2002 case was the only MSSW observed in the Southern Hemisphere, whereas the 2009 case was the strongest on record. The 1989 case was also added to this study, as Harada et al. (2010) pointed out the similarity of this case to the 2009 MSSW in terms of the role of the development of tropospheric ridges, among others. Existing studies on the predictability of SSWs mainly target vortex displacement SSWs in the Northern Hemisphere (Mukougawa and Hirooka 2004; Mukougawa et al. 2005; Hirooka et al. 2007). Stratospheric predictability in the Southern Hemisphere also remains relatively unexplored.

This paper is organized as follows. Section 2 documents the reanalysis and HC data analyzed in this study. Section 3 examines general and detailed predictability features of the MSSWs and the associated wave forcing for the pre-MSSW period. Section 4 examines the predictability connection between stratospheric and tropospheric circulation anomalies for the post-MSSW period. Section 5 provides a summary and discussion.

2. Data

a. JRA-25/JCDAS data

This study uses daily averages in the JRA-25/JCDAS data (Onogi et al. 2007). We regard the reanalysis data as a reference dataset representing the real world. The horizontal resolution of the analyzed data is 2.5° × 2.5°, with 23 pressure levels up to 0.4 hPa. The data analyzed here span 31 years from 1979 to 2009 to match the HC data described in section 2b: the JRA-25 data extend until 2004, after which the JCDAS data are available. The observed climatology and variability (standard deviation) are obtained for the 31 years. The data are referred to as the JRA-25 data for simplicity.

The zonal mean zonal wind at 60°N/S, 10 hPa in the reanalysis data confirms the occurrence of the three MSSWs, as it exhibits zonal wind reversals for the target period after the westerly wind weakens (Fig. 1). The quantity at the location is standard when one looks for the occurrence of MSSWs (e.g., Charlton and Polvani 2007). It is notable that the 2002 case experiences a more sudden weakening in the zonal wind than that observed for the two other cases.

Fig. 1.
Fig. 1.

Zonal mean zonal wind at 60°N/S, 10 hPa for the three MSSWs as a function of time lag: 2002 (solid), 2009 (dash), and 1989 (dashed–dotted). Lag-0 day corresponds to the key days when the easterly winds maximize. Black ticks denote the initial dates of the HC experiments examined in this study (see Table 1).

Citation: Journal of the Atmospheric Sciences 71, 8; 10.1175/JAS-D-13-078.1

We define key days of the three MSSWs (denoted as lag = 0 day) as when the zonal mean zonal wind at the location in the reanalysis data indicates maximum easterlies (Fig. 1). For reference, the dates are 27 September 2002, 29 January 2009, and 25 February 1989 (Table 1). We focus on the days of maximum easterlies, instead of easterly onset days, because it seems more appropriate for the comparison among the three cases. The three cases have different time lags between the days of wind reversal and the days of maximum easterlies.

Table 1.

Key days of the three MSSWs and initial dates of HC experiments examined in this study. The HC data are classified into groups A–D according to time lags (given in parentheses) of the initial dates with respect to the key days.

Table 1.

b. JMA 1-month HC data

For HC data, we utilize daily averages from HC experiments conducted by the JMA. The experiments are based on the March 2011 version of the JMA 1-month ensemble prediction system. The system employs a global atmospheric model of a horizontal resolution of TL159 (triangular truncation with a linear reduced Gaussian grid, horizontal grid size of about 110 km) and 60 vertical levels up to about 0.1 hPa.

The model includes a parameterization scheme for orographic gravity wave drag based on Iwasaki et al. (1989). The parameterization consists of two components: one for long waves (wavelengths longer than 1000 km) and the other for short waves (wavelengths around 10 km). The long waves are assumed to propagate upward until reaching wave-breaking levels mainly in the stratosphere and exert drag there. The short waves are regarded as trapped and dissipated in the lower troposphere.

The forecasts are initialized on the tenth, twentieth, and last day of each month for the 31 years from 1979 to 2009, using the JRA-25/JCDAS data. Five ensemble runs are conducted for 33 days from each initial date: the ensemble runs are different in initial perturbations, extending into the troposphere of the Northern Hemisphere and the tropical Southern Hemisphere.

For each of the three MSSWs, we examine all HC data with lead times ranging from about lag = −30 to +5 days (Table 1 and Fig. 1; black ticks in Fig. 1 denote the initial dates). The HC data are classified into four groups (A–D) according to the initial dates with respect to the key days. For example, for the 2002 MSSW, the HC A–D data have lead times of −27, −17, −7, and +3 days, respectively. Note that the exact timing of the initial dates in each group is somewhat different among the three cases. We examine the HC A–C data initialized before the key days, when we are interested in the predictability of the MSSWs. It turns out that for the three MSSWs, the HC C data generally capture the MSSWs, whereas the HC A data do not at all. The maximum lead times of about 30 days are thus sufficiently long for our purpose: HC data with longer lead times are not useful for the purpose. The HC D data initialized after the key days are added to examine tropospheric circulation anomalies after the MSSWs.

3. Predictability for the pre-MSSW period

a. General features

We examine a few dynamical measures in the reanalysis and HC data for the pre-MSSW period to show that the predictability of MSSWs is closely related to that of preceding wave properties in the lower stratosphere and troposphere.

Figures 2a–c depict the time series of the zonal mean zonal wind at 60°N/S, 10 hPa in the reanalysis (black) and HC A–C (colors) data for the three MSSWs. For the HC data, thick lines denote the ensemble means and thin lines denote the respective ensemble members. One can see that the predictability of the zonal wind variations, or the MSSWs, degrades with increasing lead time; in particular, the degradation is notable when the lead times exceed approximately 10 days. The HC A and B data with lead times of about 2 weeks or more largely depart from the reanalysis data for all cases. On the other hand, the HC C data, initialized 10 days or less before the key days, agree well with the reanalysis data, although the reproducibility varies with cases and ensembles; for example, the HC C data for the 2002 case barely qualify as an MSSW.

Fig. 2.
Fig. 2.

(a)–(c) Zonal mean zonal wind at 60°N/S, 10 hPa and (d)–(f) poleward eddy heat flux in 45°–75°N/S, 100 hPa for the reanalysis and HC data as functions of time lag. Black lines denote the reanalysis data. Color lines denote the HC data: green for the HC A group, blue for B group, and purple for C group. Thick lines denote the ensemble means, and thin lines denote respective ensemble members. Cases shown are (a),(d) 2002; (b),(e) 2009; and (c),(f) 1989.

Citation: Journal of the Atmospheric Sciences 71, 8; 10.1175/JAS-D-13-078.1

To be more specific, it is very difficult to predict the zonal wind reversals, or easterly onsets of the three vortex split MSSWs, with lead times longer than about 2 weeks (Figs. 2a–c). Some or all members of the HC C group predict the easterly onsets around the actual onset days, whereas no ensemble member of the HC A or B group demonstrates a zonal wind reversal around the onset days. One member of the HC B group for the 1989 case shows a zonal wind reversal on lag = +1 day (Fig. 2c), but the timing is late: the actual reversal occurs on lag = −4 days. The temporal variations leading to the reversals are also quite different.

A further inspection of Figs. 2a–c reveals differences among the three cases. In particular, the HC B data with lead times of about 2 weeks have different predictability features among the three cases: the HC B data for the 2009 case hardly exhibit zonal wind deceleration after an approximate lag = −10 days, whereas those for the 2002 and 1989 cases do to some extent. In other words, for the 2009 case, the HC B data are as poor as the HC A data in predicting the wind change, although the initial date of the HC B data is 10 days closer to the key day. The poor reproducibility of the 2009 MSSW by the HC B data differs from Kim and Flatau (2010), for whom some HC data initialized on 10 January (lag = −19 days in our terminology) revealed a deceleration of the polar night jet when the orographic drag scheme is improved, as mentioned in introduction.

In the real world (reanalysis data), zonal wind variations in the three MSSWs are accompanied by increases in the poleward eddy heat flux (Figs. 2d–f), as is well known (e.g., Polvani and Waugh 2004). The heat flux here includes contributions from all wave components in the daily averaged data and is averaged between 45° and 75°N/S at 100 hPa. Note that these panels plot poleward heat flux in each hemisphere as positive values: in other words, the heat flux in the Southern Hemisphere, which is usually negative, is inverted in sign. The results are independent of the exact choice of the latitudinal band for the average. The accompaniment is interpreted as the deceleration of the mean zonal wind in the stratosphere by the driving of the upward propagation of enhanced planetary waves, consistent with the dynamical model of SSWs (Matsuno 1971). The heat flux has the largest and sharpest (or most isolated) peak for the 2002 case.

The predictability of the zonal wind is closely related to that of the heat flux for the three MSSWs. The HC data that follow zonal wind weakening/reversals show increases in the heat flux (HC C) and vice versa (HC A and B). The HC B data for the 2002 and 1989 MSSWs somewhat underestimate the reanalysis heat flux: a period between lag = −10 and 0 days for example is observed. The heat flux in the HC B data for the 2009 case is significantly smaller than that in the reanalysis data. The HC C data for the 2002 case well reproduced the heat flux peak but decreased earlier: the latter feature is consistent with the result that the deceleration of the zonal wind in the HC C data is insufficient (Fig. 2a).

The close predictability relationship between zonal wind and heat flux is evident in the scatter diagrams (Fig. 3). The figure plots cumulative heat flux for lag = −20 to 0 days (x axis) and mean zonal wind for lag = −2 to +2 days (y axis). The former time window considers the sensitivity of the polar vortex to cumulative heat flux in the lower stratosphere (Polvani and Waugh 2004), whereas the latter time window seeks to capture mean conditions of the vortex around the key days. The reanalysis data are substituted before the initial dates when HC data are needed but unavailable in cumulating the heat flux. For example, of the 2002 case, the HC C data exist only after lag = −7 days (Table 1). The reanalysis data are substituted for lag = −20 to −8 days to obtain the cumulative heat flux for lag = −20 to 0 days. The figure also plots 95% confidence intervals for each quantity, obtained with Student’s t distribution for the five ensemble members.

Fig. 3.
Fig. 3.

Scatterplots of cumulative poleward eddy heat flux in 45°–75°N/S, 100 hPa (x axis) and mean zonal wind at 60°N/S, 10 hPa (y axis). The heat flux is cumulative from lag = −20 to 0 days. The zonal wind is averaged between lag = −2 and +2 days. Black squares denote the reanalysis data. Color markers denote the HC data: green for HC A group, blue for B group, and purple for C group. Larger (filled) markers denote the ensemble means, and smaller (open) markers denote the respective ensemble members. Horizontal and vertical solid lines denote 95% confidence intervals of each quantity based on the Student’s t distribution for the five members. The reanalysis data are substituted before the initial dates when HC data are needed but unavailable in cumulating the heat flux.

Citation: Journal of the Atmospheric Sciences 71, 8; 10.1175/JAS-D-13-078.1

All three panels of Fig. 3 exhibit quasi-linear distributions from top left to bottom right. This indicates negative correlations between the two quantities in terms of predictability for the three MSSWs. Note that negative correlations are notable for two perspectives: when looking at the ensemble means for different lead times and when looking at the ensemble members for the respective lead times. The negative correlations in the former perspective indicate the general improvement of predictability for both heat flux and zonal wind with decreasing lead times, although the 95% confidence intervals overlap in some cases.

For the 2002 case, the HC C data underestimate both the heat flux and the zonal wind decrease (cf. Figs. 2a,d). It is also noteworthy that the difference between the HC B and reanalysis data is much greater for the 2009 case than for the two other cases. For the 2009 case, the HC B data are located near the HC A data in the diagram (Fig. 3b).

b. In-detail features for each case

The close predictability relationship between the stratosphere and troposphere for the pre-MSSW period results from the close connection of the preceding wave activity with the stratospheric response, or MSSWs. Here we further examine dynamical features in the lower atmosphere for the pre-MSSW period to better understand predictability changes with time lag and cases. We show that the HC B data in all three cases have differences in wave properties from the reanalysis data, as diagnosed by three-dimensional wave activity flux (WAF) and geopotential height waves. Furthermore, for the 2009 case, the mean zonal wind also plays an important role.

1) 2002 Southern Hemisphere case

Figure 4a shows longitude–height sections of the WAF (arrows and gray shades) of quasi-stationary waves 1–3 for 45°–75°S of the reanalysis data for the pre-MSSW period (lag = −10 to 0 days) of the 2002 case. The geopotential height waves are also shown by black contours. The WAF is based on Plumb (1985) and Harada et al. (2010). This period is chosen when the 100-hPa poleward heat flux is large in the reanalysis data and the HC B data show large differences from the reanalysis data for all three MSSWs (Fig. 2).

Fig. 4.
Fig. 4.

Longitude–height sections of the WAF (arrows and gray shades) of quasi-stationary waves 1–3 in the (a),(c),(e) reanalysis data and (b),(d),(f) HC B data. The zonal and vertical components of the WAF for 45°–75°N/S for lag = −10 to 0 days are plotted. The WAF of the HC data is calculated from the ensemble-mean fields. Light and dark gray shades denote where the vertical component exceeds 0.04 and 0.08 m2 s−2, respectively. The quasi-stationary waves are extracted with a 5-day running mean. The geopotential height waves 1–3 are also shown by black contours for the same latitudinal and time-lag conditions. Cases shown are (a),(b) 2002; (c),(d) 2009; and (e),(f) 1989.

Citation: Journal of the Atmospheric Sciences 71, 8; 10.1175/JAS-D-13-078.1

The JRA-25 data show a large WAFz (vertical component of WAF) at 100 hPa in this period around 30°W–120°E, consistent with Fig. 2d. This packet originates from the ridge in the upper troposphere, which is located around 90°W–0° for this period. These features are pointed out by Nishii and Nakamura (2004). The HC B data broadly reproduce the enhanced WAF at 100 hPa, but somewhat underestimate the magnitude and zonal extent (Fig. 4b). The HC B data also underestimate the strength of the trough in the lower stratosphere, around 45°E. Contributions from other zonal regions are relatively small (not shown).

The underestimation in the 100-hPa WAF is also notable in time–longitude sections of 300-hPa geopotential height waves, with large 100-hPa WAFz marked by dots (Figs. 5a,d). The reanalysis results (Fig. 5a) show a region of large WAFz around 0°–120°E for about lag = −10 to 0 days (as marked by gray and white dots). It is notable that the HC B data underestimate the extent and location of the region (Fig. 5d): the region of white dots is more limited in longitude and time, while gray dots are located too far eastward. This seems related to the ridge at 300 hPa, located between 90°W and 0° and having a smaller amplitude and shorter persistence in the HC data for the period, because the large 100-hPa WAFz originates from the ridge.

Fig. 5.
Fig. 5.

Time lag–longitude sections of the quasi-stationary geopotential height waves 1–3 for 45°–75°N/S at 300 hPa. Shown are (a)–(c) the reanalysis data and (d)–(f) the ensemble means of the HC B data for cases (a),(d) 2002; (b),(e) 2009; and (c),(f) 1989. Contour interval is 100 m. Gray and white dots denote where the WAFz at 100 hPa exceeds 0.04 and 0.08 m2 s−2, respectively.

Citation: Journal of the Atmospheric Sciences 71, 8; 10.1175/JAS-D-13-078.1

The synoptic maps (Figs. 6a,d) also show some differences in the 300-hPa geopotential waves around lag = −5 days (averaged for lag = −7 to −3 days). In particular, the HC B data differ from the reanalysis data in the ridge (magnitude and zonal structure). The period for the maps is narrowed to retain time-varying features when Figs. 5a and 5d suggest the large differences between the reanalysis and HC data in the 300-hPa height and also 100-hPa WAFz. These results are consistent with Nishii and Nakamura (2004) and Allen et al. (2006), who noted the importance of the tropospheric ridge in reproducing the 2002 MSSW.

Fig. 6.
Fig. 6.

Synoptic maps of the low-passed geopotential height at 300 hPa for lag = −5 ± 2 days in (a)–(c) the reanalysis data and (d)–(f) the ensemble means of the HC B data. Black contours denote the total fields (zonal mean and all waves; contour interval is 200 m), and color shades denote waves 1–3 (contour interval is 100 m; warm and cold colors denote positive and negative values, respectively). Shown are cases (a),(d) 2002; (b),(e) 2009; and (c),(f) 1989.

Citation: Journal of the Atmospheric Sciences 71, 8; 10.1175/JAS-D-13-078.1

2) 2009 Northern Hemisphere case

The HC B data for the 2009 MSSW is characterized by the severe underestimation in the 100-hPa heat flux approximately 2 weeks prior to the key day (Fig. 2e). Harada et al. (2010) examined the WAF in detail to demonstrate that the upper-tropospheric ridge over Alaska excites wave packets propagating eastward and upward (first stage) and subsequently the packets intensify planetary waves over Siberia (second stage). That the HC B data underestimate the 100-hPa heat flux for about 2 weeks suggests that the data fail to well represent the behavior for either stage (or both stages).

It turns out that the HC data reasonably reproduce the large WAFz at 100 hPa for the first stage from about lag = −15 to −10 days around 180°–90°W, but they become more unsuccessful for the second stage from about lag = −10 to 0 days (Figs. 5b,e). Figures 4c and 4d show that the large WAFz well extends to the 100-hPa level around 60°–150°E in the reanalysis data, whereas it does not in the HC data. This is consistent with the fact that the geopotential height waves in the HC data have smaller extrema (in magnitude) of the trough and ridge in 60°E–90°W and a smaller phase tilt with height in the lower stratosphere and upper troposphere (e.g., zero contour line near 180° between 100 and 300 hPa). From Figs. 5b and 5e for the HC data, the ridge and trough at 300 hPa decay earlier around lag = −10 to −5 days to have smaller amplitudes around lag = −5 days. Synoptic maps of 300-hPa height waves confirm that the ridge and trough in the HC data are smaller in magnitude than those in the reanalysis data (Figs. 6b,e).

The mean zonal wind profile also plays an important role in determining the heat flux in the lower stratosphere for the 2009 case. Figure 7 examines the mean zonal wind and refractive index (for zonal wavenumber 2) in the reanalysis and HC data around lag = −5 days (averaged for lag = −7 to −3 days), together with the zonal mean WAF (meridional and vertical components). The refractive index is calculated for wave 2, since it is a dominant component of planetary waves in this period. It is confirmed that the upward extension of WAFz is more limited in the HC data. The reanalysis WAF exhibits marked poleward (in addition to upward) propagation around 60°N above the middle troposphere, which is not well reproduced in the HC data.

Fig. 7.
Fig. 7.

Latitude–height sections of the mean zonal wind (thick contours; m s−1) and refractive index for wave 2 (thin contours and shaded) for lag = −5 ± 2 days of the 2009 case for (a) the reanalysis data and (b) the ensemble mean of the HC B data. Light and dark shades denote where the refractive index (multiplied by the square of Earth’s radius) exceeds 80 and 160, respectively. (c) The zonal wind difference [(b) minus (a)]; contours are drawn with a 2 m s−1 interval for negative values and a 10 m s−1 interval for zero and positive values, with additional contours at +5 m s−1. (d),(e) As in (a),(b), but for zonal means of the meridional and vertical components of WAF. Light and dark shades denote where the zonal mean WAFz exceed 0.04 and 0.08 m2 s−2, respectively.

Citation: Journal of the Atmospheric Sciences 71, 8; 10.1175/JAS-D-13-078.1

These differences are consistent with distributions of the mean zonal wind and refractive index. The index of the HC data is smaller than that of the reanalysis data in the middle troposphere to lower stratosphere of high latitudes, associated with the zonal wind difference (Fig. 7c; see also Figs. 8a,c). In particular, the stronger westerly wind itself in the HC data contributes to the smaller refractive index in the middle to upper troposphere around 60°N (not shown), where the large WAF differences occur (Figs. 7d,e). The mean zonal wind profile in the HC data does not promote the upward and poleward propagation as observed, contributing to the severe underestimation of the 100-hPa heat flux (Figs. 2b,e). An inspection of individual ensemble members further supports this result: each ensemble member has stronger westerly winds in the upper troposphere of high latitudes than the reanalysis data (not shown), as well as a weaker heat flux at 100 hPa.

Fig. 8.
Fig. 8.

Time lag–latitude sections of the mean zonal wind (gray contours and shaded) and Eliassen–Palm flux divergence (black contours drawn at −12, −8, and −4 m s−1 day−1) for the (a),(c) 2009 and (b),(d) 1989 MSSWs in (a),(b) the reanalysis data and (c),(d) the ensemble means of the HC B data. Both quantities are averaged over 200–500 hPa. Light shades are applied to wind values between 0 and 10 m s−1 and dark shades are applied to those above 10 m s−1. The Eliassen–Palm flux divergence is smoothed in time and latitude for graphical purposes.

Citation: Journal of the Atmospheric Sciences 71, 8; 10.1175/JAS-D-13-078.1

This result implies that wave–mean flow interaction around the upper troposphere of high latitudes, which affects the zonal wind profile in that region, is important for good MSSW predictions. The weaker westerly wind for the reanalysis data is associated with a stronger deceleration around lag = −15 days, induced by sharper wave-driving peaks, or the convergence of the Eliassen–Palm flux (Fig. 8a). Wave driving for the HC B data is weaker and broader, consistent with stronger westerly wind (Fig. 8c). The respective ensemble members do not reproduce the observed wave driving behavior well (not shown). It is possible that the orographic gravity wave drag also plays a role in shaping such zonal wind profiles, as the drag is known to affect the mean zonal wind near the subtropical tropopause region in the Northern Hemisphere, and therefore have an impact on the propagation of planetary waves, for example, for the mean climate and its change (Sigmond and Scinocca 2010).

3) 1989 Northern Hemisphere case

Harada et al. (2010) pointed out that the 1989 MSSW is similar to the 2009 MSSW, in the sense that a ridge over Alaska plays an important role in planetary wave forcing in each case. One can confirm that the geopotential height waves and WAF are more or less similar between the two cases in the reanalysis data (Figs. 4c,e and Figs. 5b,c). The 1989 case is further characterized by the longer persistence of the ridge and trough around 120°E–110°W (Fig. 5c).

The HC B data well reproduce a region of large WAFz over the ridge until about lag = −5 days; however, the HC data do not show another over the trough after an approximate lag = −10 days (Fig. 5f). A comparison between Figs. 4e and 4f demonstrates that the WAFz in the HC B data extends less upward around 120°E–150°W. Synoptic maps of 300-hPa height demonstrate that the magnitude of the ridge is somewhat underestimated in the HC data around lag = −5 days (Figs. 6c,f).

The HC B data better reproduce the heat flux in the lower stratosphere and upper troposphere for the 1989 case than for the 2009 case (Figs. 2, 3). This may result because the difference in the mean zonal wind profile has a weaker impact for the 1989 case. For the 1989 case, there are also noticeable differences in zonal wind and wave driving between the reanalysis and HC data in the pre-MSSW period; however, the wind difference for this case is smaller than that for the 2009 case (Figs. 8b,d).

4. Predictability for the post-MSSW period

a. General features

We examine the connection of the predictability of the MSSWs to that of subsequent tropospheric anomalies. We show that the connection for the post-MSSW period (approximately 2 weeks after the key days) is more subtle than that for the pre-MSSW period, as improvements of the stratospheric and tropospheric predictability with the initial dates are often asynchronous.

Figure 9 plots the time series of the zonal mean geopotential height poleward of 65°N/S at 10 hPa and near the surface for the three MSSWs. We use the geopotential height over the polar cap to characterize extratropical circulation, especially in the troposphere, as the polar cap height is highly correlated with the northern annular mode variability (Baldwin and Thompson 2009). We applied the 850- and 1000-hPa heights as the near-surface height in the Southern and Northern Hemispheres, respectively. Anomalies of the near-surface height are related to other surface variables, such as circulation, air temperature, and precipitation (e.g., Thompson and Wallace 2000; Sigmond et al. 2013). The figure includes results from the HC D data (red lines), initialized after the key days.

Fig. 9.
Fig. 9.

As in Fig. 2, but for the zonal mean geopotential height (m) averaged poleward of 65°N/S (a)–(c) at 10 hPa and (d)–(f) near the surface. Heights are (d) 850 hPa (1100 m is subtracted for graphical purposes) and (e),(f) 1000 hPa. Results from the HC D group are added in red.

Citation: Journal of the Atmospheric Sciences 71, 8; 10.1175/JAS-D-13-078.1

The stratospheric portion (Figs. 9a–c) well reproduces the MSSW predictability result in the Figs. 2a–c. This is expected, since the geopotential height is strongly related to zonal wind by the geostrophic wind relationship. On the other hand, such predictability changes for the near-surface height after the MSSWs are not systematically clear. For example, the HC C data reproduce the reanalysis 10-hPa height much better than the HC B data for all three cases, whereas such improvements are unclear in the near-surface height a few weeks after the key days. Even the HC D data, initialized after the key days, do not always lead to a better reproducibility after the MSSWs, compared to the HC C data.

Figures 10a–c display scatterplots of normalized geopotential height anomalies over the polar cap at 10 hPa and near the surface. The 10-hPa height is averaged from lag = −2 to +2 days, and the near-surface height is from lag = +1 to + 15 days. The figure excludes the HC A data due to insufficient data after the MSSWs. Note that both axes use normalized height anomalies by the climatological mean and standard deviation (of interannual variability) of the reanalysis data. The anomalies, even when normalized, have greater magnitudes in the stratosphere than in the troposphere.

Fig. 10.
Fig. 10.

As in Fig. 3, but for the normalized zonal mean geopotential height anomalies poleward of 65°N/S. The x axis uses the 10-hPa height for (a)–(c) lag = −2 to +2 days and (d)–(f) lag = +1 to +15 days. The y axis uses near-surface height for lag = +1 to +15 days: (a),(d) 850 and (b),(c),(e),(f) 1000 hPa. The results are colored as in Fig. 9. The HC A data are excluded, owing to insufficient data after the MSSWs. The reanalysis data are substituted before the initial dates, when HC data are needed but unavailable in calculating the time means, as in Fig. 3. The same treatment is also applied in Figs. 12 and 13.

Citation: Journal of the Atmospheric Sciences 71, 8; 10.1175/JAS-D-13-078.1

The plots confirm for the 2002 and 1989 cases that although the ensemble means of the HC data generally approach the JRA-25 data, as initialized later, this is more notable at 10 hPa (in the horizontal direction in the figure) than near the surface (in the vertical direction). In other words, the improved predictability of the near-surface height with the initial dates is much weaker than that at 10 hPa. The 95% confidence intervals of near-surface height anomalies tend to overlap for the different initial dates, whereas the 10-hPa counterparts are often separate. This feature holds when using a longer averaging period of lag = +1 to +15 days for the 10-hPa height (Figs. 10d–f). These results for the post-MSSW period are in contrast to those for the pre-MSSW period (Fig. 3). For the 2009 case, on the other hand, the HC data approach the JRA-25 data both at 10 hPa and near the surface, as initialized later (Figs. 10b,e).

b. In-detail features for each case

Since the downward propagation of circulation anomalies is a salient feature in the composite picture after SSWs (e.g., Baldwin and Dunkerton 2001), time–height sections of the normalized zonal mean geopotential height over the polar cap are shown for the JRA-25 and HC B-D data in Fig. 11. The figure also plots gray dots for statistically significant differences of the HC data from the reanalysis data at the 95% level. The significance is obtained when the 95% confidence interval of the HC data determined with the Student’s t distribution for the five ensemble members is different from the JRA-25 data. It is notable that while the stratospheric height above about 100 hPa around the key days tends to be better reproduced in the HC data by delaying the initial dates, such a tendency is relatively unclear for tropospheric height. In other words, the HC data do not exhibit the vertical coupling after the MSSWs seen in the reanalysis data, even when they well reproduce (or include) the observed stratospheric anomalies around the key days. Each case is examined as follows.

Fig. 11.
Fig. 11.

Time–height sections of the normalized zonal mean geopotential height anomalies averaged poleward of 65°N/S for the three cases: (a),(d),(g),(j) 2002; (b),(e),(h),(k) 2009; and (c),(f),(i),(l) 1989 for (a)–(c) the reanalysis data and the ensemble means of (d)–(f) HC B, (g)–(i) HC C, and (j)–(l) HC D. Contour interval is 0.5. Gray dots in (d)–(l) indicate that the differences of the HC data from the JRA-25 data are statistically significant at the 95% confidence level (see the text for details). Gray dots are marked only where the differences of the HC ensemble mean from the JRA-25 data exceed ±0.5.

Citation: Journal of the Atmospheric Sciences 71, 8; 10.1175/JAS-D-13-078.1

1) 2002 Southern Hemisphere case

Geopotential height anomalies in high latitudes after the 2002 MSSW are positive in both the stratosphere and troposphere, with much larger magnitudes in the former (Fig. 11a). The stratospheric feature is more or less reproduced in the HC B–D data, whereas the tropospheric counterpart is not (Fig. 11, left): the HC data include negative height anomalies in the troposphere. This is consistent with the synoptic maps of 850-hPa height anomalies (Fig. 12, left). The reanalysis data have predominantly positive anomalies at high latitudes; however, the HC data, in particular B and C, underestimate height anomalies.

Fig. 12.
Fig. 12.

Synoptic maps of the near-surface geopotential height for lag = +1 to +15 days of the three MSSWs: (a),(d),(g),(j) 2002 case at 850 hPa, (b),(e),(h),(k) 2009 at 1000 hPa, and (c),(f),(i),(l) 1989 at 1000 hPa for (a)–(c) the reanalysis data and the ensemble means of (d)–(f) HC B, (g)–(i) HC C, and (j)–(l) HC D. Colors denote anomalies from the climatology of the reanalysis data (contour interval is 50 m). Black contours denote differences of the HC data from the reanalysis data (contour interval is 50 m). Coastal lines (white) and reference longitudes are drawn in (a)–(c).

Citation: Journal of the Atmospheric Sciences 71, 8; 10.1175/JAS-D-13-078.1

Such reproducibility is quantified by calculating anomaly correlations of geopotential height anomalies. The quantification is based on the spatial correlation for time-averaged height anomalies (poleward of 20°N/S) at each level between the reanalysis and each HC member. The anomalies are defined as those from the climatology of the reanalysis data. The time average is taken from lag = +1 to +15 days. Confidence intervals at the 95% level are obtained with the results from the five members, and depicted by horizontal lines in Fig. 13a. The anomaly correlations are also calculated for the ensemble mean fields. This method is reasonable to examine the reproducibility of height anomalies over middle and high latitudes.

Fig. 13.
Fig. 13.

Vertical profiles of anomaly correlations between the reanalysis and HC B–D data for the three MSSWs: (a) 2002, (b) 2009, and (c) 1989. The results are colored as in Fig. 9. The anomaly correlations are calculated for geopotential height anomalies poleward of 20°N/S for lag = +1 to +15 days between the reanalysis and each ensemble member, and 95% confidence intervals are obtained with Student’s t distribution (denoted by horizontal lines at each level). The horizontal lines are slightly offset for graphical purposes. The markers (triangles, circles, and diamonds) denote the anomaly correlations of the ensemble mean fields.

Citation: Journal of the Atmospheric Sciences 71, 8; 10.1175/JAS-D-13-078.1

A common feature among the HC B–D data is that the anomaly correlations are higher in the stratosphere (close to 1 for the HC C and D data) and lower in the troposphere (Fig. 13a). This feature is consistent with Fig. 10d. The anomaly correlations in the troposphere are about 0.25 (when averaged between 200 and 850 hPa) for the HC B and C data, and they increase to about 0.7 for the HC D data. Notable differences in 850-hPa height anomalies exist even for the HC D data (Fig. 12j): the HC D data underestimate the strength of positive anomalies near the Antarctic Peninsula and produce a maximum over the southern Indian Ocean, which is not observed.

It is also notable that changes in the anomaly correlations with HC groups are asynchronous between the stratosphere and troposphere (Fig. 13a). The correlations in the stratosphere increase from the HC B to C data, but those in the troposphere do not. The situation is different between the HC C and D data; both have very high correlations in the stratosphere, but their correlations increase in the troposphere.

2) 2009 Northern Hemisphere case

The height anomalies observed after the 2009 MSSW exhibit the familiar downward propagation (values over +2) in the stratosphere (Fig. 11b). Positive anomalies are also present in the troposphere until about lag = +25 days. The 1000-hPa height anomalies are characterized by positive anomalies over the pole and the North Pacific (Fig. 12b).

The HC B data substantially vary from the reanalysis data in both the stratosphere and troposphere, not depicting any positive anomalies at all (Figs. 11e and 12e). The anomaly correlations are low, around 0 in the troposphere or even negative in the stratosphere (Fig. 13b). The HC C data marginally reproduce the stratospheric anomalies around the key days, but they lack the downward-propagation feature and positive anomalies in the troposphere (Figs. 11h, 12h). This is consistent with higher anomaly correlations in the stratosphere and lower correlations in the troposphere (Fig. 13b; see also Fig. 12h). Note also that changes in the anomaly correlations from the HC B to C data are different between the stratosphere and troposphere: correlations in the stratosphere notably improve, whereas such improvements are much weaker in the troposphere.

The HC D data, initialized after the key day, best reproduce the stratospheric and tropospheric anomalies as expected, although they fail to show the sign reversal of the tropospheric anomalies around lag = +25 days (Fig. 11k). The 1000-hPa height anomalies in the HC D data resemble those in the reanalysis data (Fig. 12k), consistent with the high correlation in the troposphere (Fig. 13b).

A comparison between the HC C and D data results in Fig. 13b reveals that predictability after the 2009 MSSW differs between the stratosphere and troposphere. The middle-stratospheric predictability is very high for both the HC C and D data, as the anomaly correlations are similar to 1 at 10 and 30 hPa for both groups. On the other hand, the tropospheric predictability significantly differs between the HC C and D data.

3) 1989 Northern Hemisphere case

The evolutions of the observed height anomalies are different between the 2009 and 1989 MSSWs (Figs. 11b,c). For the 1989 case, the stratospheric anomalies are smaller in magnitude and the tropospheric anomalies are predominantly negative.

Nonetheless, the different predictability features between the stratosphere and troposphere for the 2009 case also hold for the 1989 case. The stratospheric predictability notably improves from the HC B to C data, whereas the tropospheric counterpart does not (Fig. 11, right; Fig. 12, right; Fig. 13c). The anomaly correlations in the middle stratosphere are very high for both the HC C and D data, whereas those in the troposphere increase from the HC C to D data. Note also that the anomaly correlations in the troposphere are lower than those in the stratosphere for the HC C and D data (this is especially the case for the former).

5. Summary and discussion

Comparing JMA 1-month ensemble HC data to JRA-25/JCDAS data, this study investigated the predictability of three MSSWs of the vortex split type (2002 Southern Hemisphere case, and 2009 and 1989 Northern Hemisphere cases) especially in two aspects: the characteristic lead time for successful predictions of the MSSWs (more generally, changes in the predictability of the MSSWs with lead time) and the connection of the MSSW predictability relative to lower-atmospheric features. Although these aspects were partly examined in existing studies (as introduced in section 1), this study has clarified more concrete, dynamical features of the MSSW predictability for the three extreme cases by using the HC data that are reasonable for the analysis. This is a pilot case study of the relatively recent extreme MSSWs, before a more statistics-oriented analysis (Taguchi 2014), using the HC data.

a. Pre-MSSW period

Among the three cases, the predictability of the MSSWs generally improves with decreasing lead times of the HC data, as is common and expected. The characteristic time scale of the predictability changes is about 10 days–2 weeks: the HC C data with lead times of about 10 days or less successfully predict the MSSWs, whereas the predictability largely degrades for longer lead times. It turns out to be very difficult to predict the MSSWs (10-hPa zonal wind reversals) with lead times of about 2 weeks or longer: no ensemble member for the condition predicts wind reversals by the actual onset days.

We demonstrated that such predictability changes for MSSWs are strongly connected to planetary wave forcings in the lower atmosphere. The high predictability of the MSSWs is accompanied by enhanced wave activity in the lower stratosphere and anomalous wave structures in the troposphere. The HC B data with lead times of 15–19 days fail to well reproduce the structure (strength and location) and persistence of ridges (and associated troughs), and hence underestimate planetary wave activity entering the stratosphere for all three cases. Taguchi (2014) finds that the HC B data for the 2009 MSSW is the most notable outlier in the spread–error relationship for the winter stratosphere in the HC data.

It is also shown that whereas the 2009 and 1989 cases are similar in observed wave forcing (Harada et al. 2010), the former is much more difficult for the HC B data to predict. We suggest that the wave–mean flow interaction in the upper troposphere of high latitudes occurring over the intraseasonal time scale, which affects wave activity entering the stratosphere (Shiotani 1986; Chen and Robinson 1992), is complicated and difficult to reproduce. The HC data underestimate the refractive index (reflecting a too-strong westerly wind), as well as the upward and poleward propagation of planetary waves: these changes in the basic state and wave activity are consistent with each other. The wave forcing for the 1989 MSSW is characterized by the more conspicuous persistence of the tropospheric ridge: this feature is a key to successful prediction of the MSSW.

A motivation for this study is to investigate the predictability of vortex split MSSWs, since most cases examined by existing case studies are vortex displacement events (see introduction). We discuss the present results (time scale and connection to lower-atmospheric features) in comparison to existing results.

The characteristic time scale for distinguishing successful and unsuccessful MSSW predictions shown in this study roughly corresponds to the time scale when planetary waves intensify in association with tropospheric disturbances (such as ridges or blocking events) and propagate to the stratosphere. It is known that tropospheric blocking events are difficult to predict (e.g., Tibaldi and Molteni 1990). The blocking events are also a barrier for good predictions of the MSSWs in this study, as Allen et al. (2006) suggested for the 2002 case.

The time scale for successful predictions in this study is apparently shorter than or comparable to the results in existing studies. The existing results range, for example, from 6 days for the Southern Hemisphere MSSW in 2002 (Allen et al. 2006), via 2 weeks for the MSSW in December 2001 (Mukougawa et al. 2005; Hirooka et al. 2007) and 9–15 days for the four MSSWs (Marshall and Scaife 2010), to 1 month for the MSSW in December 1998 (Mukougawa and Hirooka 2004). Such differences reflect several factors. First, different prediction systems may have different prediction characteristics or skills. Next, each study (or experiment) has varying constraints—for example, in ensemble size and time resolution (frequency of initializations)—and employs different definitions for SSWs and successful predictions. Finally, predictability may vary for different SSW events, as suggested from the HC B data in this study (Fig. 2). The shorter time scale for the vortex split MSSWs may be a manifestation of the different time scales between waves 1 and 2 (shorter for wave 2; Hirota and Sato 1969; Mukougawa and Hirooka 2004). Whereas the present results seemingly fall in the range of the previous studies, the method and analysis of this study are more reasonable than those of some previous studies: we use the well-known objective definition for MSSWs and compare the three extreme MSSWs in the HC data.

This study generally agrees with existing studies in pointing out the importance of generation and propagation of planetary waves in the lower atmosphere, although their details and relative contributions are different among the cases examined. For example, Mukougawa and Hirooka (2004) claimed the importance of wave-1 generation near the surface for the December 1998 MSSW, whereas Mukougawa et al. (2005) drew an attention to the role of mean zonal wind in the upper troposphere for the December 2001 MSSW. Further detailed comparisons are beyond the scope of this study and are left to a future work. The importance of tropospheric planetary wave forcing is also consistent with idealized model experiments by Sun et al. (2012).

Our results for the 2002 case agree well with Allen et al. (2006) and Nishii and Nakamura (2004) with respect to the importance of wave forcing emanating from the South Atlantic ridge. We further show from the HC C data that the persistence of the ridge affects the strength of the cumulative heat flux and MSSW (easterly wind). It is also common between Kim and Flatau (2010) and this study for the 2009 MSSW that the deceleration of the westerly wind in the stratosphere is underestimated when the eddy heat flux in the lower stratosphere is underestimated. Our results further show that the underestimation in the heat flux is also associated with the strength and persistence of the upper-tropospheric ridge and trough in this case.

Kim and Flatau (2010) demonstrate that the 2009 MSSW is predictable in some experiments with lead times of about 2 weeks. This is longer than our result, in the sense that no ensemble member predicts a zonal wind reversal in the HC B and C data with similar lead times (Fig. 2b). Kim and Flatau suggest the important role of the parameterized orographic gravity wave drag in decelerating the mean westerly wind in the extratropical middle stratosphere and above in their experiments. The role of the gravity wave drag can be a candidate for explaining the difference, since the different parameterization schemes are used in the two experiments and are likely to make different contributions.

b. Post-MSSW period

The connection of the predictability between stratospheric and tropospheric anomalies is more subtle for the post-MSSW period (about 2 weeks after the MSSWs) than for the pre-SSW period. First, the predictability of post-MSSW stratospheric and tropospheric anomalies is different within the same HC group: in particular, the HC C group initialized about 1 week before the MSSWs tends to reproduce the evolution of the stratosphere after the MSSWs well (characterized by higher anomaly correlations) but not that of the troposphere. Second, the predictability is generally improved in both the stratosphere and troposphere with delaying initial dates, but the improvements are often asynchronous between the two regions, particularly for the HC B and C groups. The tropospheric predictability sometimes remains similar from one HC group to the next, while the stratospheric predictability improves, or vice versa.

The subtle predictability connection (i.e., smaller anomaly correlations in the troposphere for the HC C group) is likely to reflect a difference in characteristic time scales for good predictions between the stratosphere and troposphere. The time scale of the stratospheric variability after the MSSWs is much longer than that of the tropospheric predictability, which is likely to reflect baroclinic eddy activity. Thus, the tropospheric anomalies are more difficult for the HC C group to predict than stratospheric anomalies in the post-MSSW period of the 2-week time window.

The subtle connection also suggests that a large fraction of intraseasonal tropospheric variability is inherently tropospheric; however, some fraction is influenced by stratospheric anomalies. This may be related to the fact that not all MSSWs are followed by a clear signal of downward propagation to the troposphere (Nakagawa and Yamazaki 2006). The predictability of the tropospheric circulation is improved after disturbed stratospheric conditions, or SSWs, in a statistical sense (Mukougawa et al. 2009; Sigmond et al. 2013), whereas such improved predictability may further vary from one case to another, for example, with conditions of downward propagation. Furthermore, predictability of tropospheric anomalies after MSSWs may have geographical dependence, as Woollings et al. (2010) show the regional anomalies in the troposphere following zonal mean flow anomalies in the stratosphere.

It is also noted that the present result for the post-MSSW period may be of limited use due to a few factors. First, the time window examined here may not be sufficiently long (due to the experimental constraint) to fully extract a possible connection, if any, after the MSSWs. Sigmond et al. (2013) used a 16–60-day window to obtain the enhanced forecast skill of tropospheric circulation after 20 SSWs, when they initialized the forecasts on the day of the SSWs. Second, such a vertical connection of predictability changes may vary with cases examined, as well as with prediction systems. These questions can be examined in future studies.

At the end of this section, it is left to future research to conduct and analyze HC experiments with more frequent initializations for a longer time scale using more samples. Such research includes our statistical analysis investigating stratospheric predictability in more general terms using the HC data (Taguchi 2014), which can be regarded as a complementary study to the present case study. We are also examining the variability of the MSSW predictability using more samples. Furthermore, the SNAP activity is such an attempt using multiple systems (Charlton and Jackson 2012): it will be very useful to improving our understanding, if a clean comparison is made using experiments, data, and analyses that are designed in a systematic manner.

Acknowledgments

The author thanks those who made the analyzed data available: JMA and the Central Research Institute of Electric Power Industry for the JRA-25/JCDAS data and the Climate Prediction Division of the JMA for the HC data. The HC data were provided by way of the “Meteorological Research Consortium,” a framework for research cooperation of the JMA and the Meteorological Society of Japan. This study is supported by the Grant-in-Aid for Scientific Research (S) 2422401101. The author also thanks the anonymous reviewers, whose comments helped to improve the manuscript.

REFERENCES

  • Allen, D. R., , L. Coy, , S. D. Eckermann, , J. P. McCormack, , G. L. Manney, , T. F. Hogan, , and Y.-J. Kim, 2006: NOGAPS-ALPHA simulations of the 2002 Southern Hemisphere stratospheric major warming. Mon. Wea. Rev., 134, 498518, doi:10.1175/MWR3086.1.

    • Search Google Scholar
    • Export Citation
  • Andrews, D. G., , J. R. Holton, , and C. B. Leovy, 1987: Middle Atmosphere Dynamics. International Geophysics Series, Vol. 40, Academic Press, 489 pp.

  • Baldwin, M. P., , and T. J. Dunkerton, 1999: Propagation of the Arctic Oscillation from the stratosphere to the troposphere. J. Geophys. Res., 104, 30 93730 946, doi:10.1029/1999JD900445.

    • Search Google Scholar
    • Export Citation
  • Baldwin, M. P., , and T. J. Dunkerton, 2001: Stratospheric harbingers of anomalous weather regimes. Science, 294, 581584, doi:10.1126/science.1063315.

    • Search Google Scholar
    • Export Citation
  • Baldwin, M. P., , and D. W. J. Thompson, 2009: A critical comparison of stratosphere–troposphere coupling indices. Quart. J. Roy. Meteor. Soc., 135, 16611672, doi:10.1002/qj.479.

    • Search Google Scholar
    • Export Citation
  • Baldwin, M. P., , T. Hirooka, , A. O’Neill, , S. Yoden, , A. J. Charlton, , Y. Hio, , W. A. Lahoz, , and A. Mori, 2003a: Major stratospheric warming in the Southern Hemisphere in 2002: Dynamical aspects of the ozone hole split. SPARC Newsletter, No. 20, SPARC Office, Verrières-le-Buisson, France, 2226.

  • Baldwin, M. P., , D. B. Stephenson, , D. W. J. Thompson, , T. J. Dunkerton, , A. J. Charlton, , and A. O’Neill, 2003b: Stratospheric memory and extended-range weather forecasts. Science, 301, 636640, doi:10.1126/science.1087143.

    • Search Google Scholar
    • Export Citation
  • Charlton, A. J., , and L. M. Polvani, 2007: A new look at stratospheric sudden warmings. Part I. Climatology and modeling benchmarks. J. Climate, 20, 449469, doi:10.1175/JCLI3996.1.

    • Search Google Scholar
    • Export Citation
  • Charlton, A. J., , and D. Jackson, 2012: SNAP: The Stratospheric Network for the Assessment of Predictability. SPARC Newsletter, No. 39, SPARC Office, Zurich, Switzerland, 4042.

  • Chen, P., , and W. A. Robinson, 1992: Propagation of planetary waves between the troposphere and stratosphere. J. Atmos. Sci., 49, 25332545, doi:10.1175/1520-0469(1992)049<2533:POPWBT>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Harada, Y., , A. Goto, , H. Hasegawa, , N. Fujikawa, , H. Naoe, , and T. Hirooka, 2010: A major stratospheric sudden warming event in January 2009. J. Atmos. Sci., 67, 20522069, doi:10.1175/2009JAS3320.1.

    • Search Google Scholar
    • Export Citation
  • Hirooka, T., , T. Ichimaru, , and H. Mukougawa, 2007: Predictability of stratospheric sudden warmings as inferred from ensemble forecast data: Intercomparison of 2001/02 and 2003/04 winters. J. Meteor. Soc. Japan, 85, 919925, doi:10.2151/jmsj.85.919.

    • Search Google Scholar
    • Export Citation
  • Hirota, I., , and Y. Sato, 1969: Periodic variation of the winter stratospheric circulation and intermittent vertical propagation of planetary waves. J. Meteor. Soc. Japan, 47, 390402.

    • Search Google Scholar
    • Export Citation
  • Iwasaki, T., , S. Yamada, , and K. Tada, 1989: A parameterization scheme of orographic gravity wave drag with two different vertical partitionings. Part I: Impacts on medium-range forecasts. J. Meteor. Soc. Japan, 67, 1127.

    • Search Google Scholar
    • Export Citation
  • Kim, Y.-J., , and M. Flatau, 2010: Hindcasting the January 2009 Arctic sudden stratospheric warming and its influence on the Arctic Oscillation with unified parameterization of orographic drag in NOGAPS. Part I: Extended-range stand-alone forecast. Wea. Forecasting, 25, 16281644, doi:10.1175/2010WAF2222421.1.

    • Search Google Scholar
    • Export Citation
  • Labitzke, K. G., , and H. van Loon, 1999: The Stratosphere: Phenomena, History, and Relevance. Springer, 179 pp.

  • Lee, J. N., , D. L. Wu, , G. L. Manney, , and M. J. Schwartz, 2009: Aura Microwave Limb Sounder observations of the northern annular mode: From the mesosphere to the upper troposphere. Geophys. Res. Lett., 36, L20807, doi:10.1029/2009GL040678.

    • Search Google Scholar
    • Export Citation
  • Limpasuvan, V., , D. W. J. Thompson, , and D. L. Hartmann, 2004: The life cycle of Northern Hemisphere sudden stratospheric warmings. J. Climate, 17, 25842596, doi:10.1175/1520-0442(2004)017<2584:TLCOTN>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Manney, G. L., and Coauthors, 2005: Simulations of dynamics and transport during the September 2002 Antarctic major warming. J. Atmos. Sci., 62, 690707, doi:10.1175/JAS-3313.1.

    • Search Google Scholar
    • Export Citation
  • Manney, G. L., and Coauthors, 2009: Aura Microwave Limb Sounder observations of dynamics and transport during the record-breaking 2009 Arctic stratospheric major warming. Geophys. Res. Lett., 36, L12815, doi:10.1029/2009GL038586.

    • Search Google Scholar
    • Export Citation
  • Marshall, A. G., , and A. A. Scaife, 2010: Improved predictability of stratospheric sudden warming events in an atmospheric general circulation model with enhanced stratospheric resolution. J. Geophys. Res., 115, D16114, doi:10.1029/2009JD012643.

    • Search Google Scholar
    • Export Citation
  • Matsuno, T., 1971: A dynamical model of the stratospheric sudden warming. J. Atmos. Sci., 28, 14791494, doi:10.1175/1520-0469(1971)028<1479:ADMOTS>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Mukougawa, H., , and T. Hirooka, 2004: Predictability of stratospheric sudden warming: A case study for 1998/99 winter. Mon. Wea. Rev., 132, 17641776, doi:10.1175/1520-0493(2004)132<1764:POSSWA>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Mukougawa, H., , H. Sakai, , and T. Hirooka, 2005: High sensitivity to the initial condition for the prediction of stratospheric sudden warming. Geophys. Res. Lett., 32, L17806, doi:10.1029/2005GL022909.

    • Search Google Scholar
    • Export Citation
  • Mukougawa, H., , T. Hirooka, , and Y. Kuroda, 2009: Influence of stratospheric circulation on the predictability of the tropospheric northern annular mode. Geophys. Res. Lett., 36, L08814, doi:10.1029/2008GL037127.

    • Search Google Scholar
    • Export Citation
  • Nakagawa, K. I., , and K. Yamazaki, 2006: What kind of stratospheric sudden warming propagates to the troposphere? Geophys. Res. Lett., 33, L04801, doi:10.1029/2005GL024784.

    • Search Google Scholar
    • Export Citation
  • Nishii, K., , and H. Nakamura, 2004: Tropospheric influence on the diminished Antarctic ozone hole in September 2002. Geophys. Res. Lett., 31, L16103, doi:10.1029/2004GL019532.

    • Search Google Scholar
    • Export Citation
  • O’Neill, A., 2003: Stratospheric sudden warmings. Encyclopedia of Atmospheric Sciences, J. R. Holton, J. A. Curry, and J. A. Pyle, Eds., Vol. 4, Academic Press, 1342–1353.

  • Onogi, K., and Coauthors, 2007: The JRA-25 Reanalysis. J. Meteor. Soc. Japan, 85, 369432, doi:10.2151/jmsj.85.369.

  • Plumb, R. A., 1985: On the three-dimensional propagation of stationary waves. J. Atmos. Sci., 42, 217229, doi:10.1175/1520-0469(1985)042<0217:OTTDPO>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Polvani, L. M., , and D. W. Waugh, 2004: Upward wave activity flux as a precursor to extreme stratospheric events and subsequent anomalous surface weather regimes. J. Climate, 17, 35483554, doi:10.1175/1520-0442(2004)017<3548:UWAFAA>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Shiotani, M., 1986: Planetary wave activity in the troposphere and stratosphere during the Northern Hemisphere winter. J. Atmos. Sci., 43, 32003209, doi:10.1175/1520-0469(1986)043<3200:PWAITT>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Sigmond, M., , and J. F. Scinocca, 2010: The influence of basic state on the Northern Hemisphere circulation response to climate change. J. Climate, 23, 14341446, doi:10.1175/2009JCLI3167.1.

    • Search Google Scholar
    • Export Citation
  • Sigmond, M., , J. F. Scinocca, , V. V. Kharin, , and T. G. Shepherd, 2013: Enhanced seasonal forecast skill following stratospheric sudden warmings. Nat. Geosci., 6, 98102, doi:10.1038/ngeo1698.

    • Search Google Scholar
    • Export Citation
  • Simmons, A., , M. Hortal, , G. Kelly, , A. McNally, , A. Untch, , and S. Uppala, 2005: ECMWF analyses and forecasts of stratospheric winter polar vortex breakup: September 2002 in the Southern Hemisphere and related events. J. Atmos. Sci., 62, 668689, doi:10.1175/JAS-3322.1.

    • Search Google Scholar
    • Export Citation
  • Stan, C., , and D. M. Straus, 2009: Stratospheric predictability and sudden stratospheric warming events. J. Geophys. Res., 114, D12103, doi:10.1029/2008JD011277.

    • Search Google Scholar
    • Export Citation
  • Sun, L., , W. A. Robinson, , and G. Chen, 2012: The predictability of stratospheric warming events: More from the troposphere or the stratosphere? J. Atmos. Sci., 69, 768783, doi:10.1175/JAS-D-11-0144.1.

    • Search Google Scholar
    • Export Citation
  • Taguchi, M., 2014: Stratospheric predictability: Basic characteristics in JMA 1-month hindcast experiments for 1979–2009. J. Atmos. Sci., doi:10.1175/JAS-D-13-0295.1, in press.

    • Search Google Scholar
    • Export Citation
  • Thompson, D. W. J., , and J. M. Wallace, 1998: The Arctic Oscillation signature in the wintertime geopotential height and temperature fields. Geophys. Res. Lett., 25, 12971300, doi:10.1029/98GL00950.

    • Search Google Scholar
    • Export Citation
  • Thompson, D. W. J., , and J. M. Wallace, 2000: Annular modes in the extratropical circulation. Part I: Month-to-month variability. J. Climate, 13, 10001016, doi:10.1175/1520-0442(2000)013<1000:AMITEC>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Tibaldi, S., , and F. Molteni, 1990: On the operational predictability of blocking. Tellus, 42A, 343365, doi:10.1034/j.1600-0870.1990.t01-2-00003.x.

    • Search Google Scholar
    • Export Citation
  • Woollings, T., , A. Charlton-Perez, , S. Ineson, , A. G. Marshall, , and G. Masato, 2010: Associations between stratospheric variability and tropospheric blocking. J. Geophys. Res., 115, D06108, doi:10.1029/2009JD012742.

    • Search Google Scholar
    • Export Citation
  • Yoden, S., , T. Yamaga, , S. Pawson, , and U. Langematz, 1999: A composite analysis of the stratospheric sudden warmings simulated in a perpetual January integration of the Berlin TSM GCM. J. Meteor. Soc. Japan, 77, 431445.

    • Search Google Scholar
    • Export Citation
Save