1. WCRP Workshop on Seasonal Prediction
Barcelona World Trade Centre, Barcelona, Spain, June 5, 2007
Predictability of Stratosphere-Troposphere Coupling during Stratospheric Sudden Warming Events in the Northern Hemisphere
Mukougawa, Hitoshi1, Yuhji Kuroda2 and Toshihiko Hirooka3
1Disaster Prevention Research Institute, Kyoto University, JAPAN 2Meteorological Research Institute, JMA, JAPAN
3Department of Earth and Planetary Sciences, Kyushu University, JAPAN
Introduction
Stratospheric Sudden Warming (SSW) events
Winter Polar Vortex
SSW H L
breakdown
NAM
Stratosphere L
(Northern Annular Mode)
time
Troposphere
Large-amplitude Planetary waves
Circulation change
(AO; Arctic Oscillation)

2. Aim of our study (1)
Examine practical predictability of stratosphere-troposphere
dynamical coupling during SSW events using operational 1-month ensemble forecast datasets provided by the JMA
(a) Predictability and tropospheric precursor for a SSW event occurring in Dec. ’01
The SSW is predictable at least 2 weeks in advance
High sensitivity of the prediction skill of the SSW to the initial condition
　　　Mukougawa et al.(2005) and Mukougawa et al. (2007)
cf. High sensitivity was also observed during the onset period of tropospheric blocking (Kimoto, Mukougawa and Yoden, 1992)
Persistence of Atlantic blocking
confirmed by a series of forecast experiments using MRI/JMA AGCM (～JMA NWP model)

3. Aim of our study (2)
(b) Stratospheric influence on the forecast skill of the tropospheric circulation Influence of stratospheric NAM variation on the forecast skill of tropospheric NAM index
5-winter (2001/ /06; DJF) archive of the JMA month ensemble forecasts
Forecast skill of upper tropospheric NAM index for a lead time from 6 to 12 days tends to be better when the negative NAM index (easterly wind anomaly) is observed in the stratosphere at the initial time of forecast
→ a better forecast skill of the tropospheric NAM index after SSW events

4. Data 1-month forecast dataset of JMA ensemble prediction system
Resolution T106L40 hybrid coordinate
Top Boundary hPa
Ozone specified by zonal-mean climatological value
SST constant SST anomaly at the
initial time
Integration Period days
Number of Ensemble members
Perturbation Method BGM(Breeding of Growing Mode)
Initialization Date Every Wednesday and Thursday
Interval of Stored Data Daily (2.5°×2.5°)
JMA Global Analysis (GANAL) dataset (1.25°×1.25°)
Verification for the prediction of JMA

5. Zonal Mean Temp. (K) ’01/’02 Winter (10hPa, 80N)
・SSW in late Dec. ’01 was caused by WN1
・Large spread for the prediction from Dec. 5 and 6
High sensitivity to initial
condition
・After Dec. 12, all ensemble
members predict SSW, and
spread becomes small
・Difference in upward
propagation of WN1 activity
among forecasts from
Dec. 5 & 6 becomes evident
around Dec. 13
Observation

6. Predicted 3-day averaged Z10 of 27-29 Dec.
Prediction from Dec. 5 & 6
180 90
270
Observation
Run 1 (S)
Run 2 (F)
Amplification of WN1 component is evident for the observation and Run 1 (the best forecast) except for the phase difference.
For Run2 (the worst forecast), robust polar vortex still exists without WN1 amplification

7. 3-day averaged Z300 Observation Run 2 (F) Initial Dec. 6-8 Onset
Prediction from Dec. 5 & 6
Observation
Run 1 (S)
Run 2 (F)
Initial
Dec. 6-8
270 90
180
Onset
period
Dec.12-14
・In Run 2, persistence of blocking over Europe is weak

8. Zonal-Mean Zonal Wind (m/s) during onset period of Dec
Zonal-Mean Zonal Wind (m/s) during onset period of Dec , 2001 (3-day mean)
Prediction from Dec. 5 and 6
hPa
Observation
Run 1 (Succeeded)
Run 2 (Failed) 10
100
1000
20N 40N 60N 80N
・For Run 2, weak easterly prevails around 80N, and strong vertical westerly shear is seen in the lower stratosphere around 60N, in common with ’98 Dec. SSW (Mukougawa and Hirooka 2004)
・High-latitude westerlies in Obs. and Run 1 are associated with blocking over Europe

9. Regressed Z300 on Dec. 12-14 with respect to Zonal-Mean Temp (10hPa, 80N) on Dec. 28
Prediction from Dec. 5 and 6
Anomaly from
Ensemble Mean
Ensemble Mean
Contour: 20m
Shaded region:Significance level > 99(95)%
・Positive height anomaly over Atlantic associated with blocking
is well correlated with the occurrence of the SSW

10. Regressed Zonal-Mean Zonal Wind on Dec. 12-14 w. r. t. Zonal-Mean Temp
Regressed Zonal-Mean Zonal Wind on Dec w.r.t. Zonal-Mean Temp. (10hPa, 80N) on Dec. 28
Prediction from Dec. 5 and 6
Anomaly from
Ensemble Mean
Ensemble Mean
1.3X105
1.3X104
5.0X107
5.0X106 (Kｇ/s2)
Contour: 5m/s
Contour: 0.5m/s
Shaded region:Significance level > 99(95)%
Arrow: WN1 EP-flux
・Westerly anomaly around 80N and easterly anomaly around
60N are well correlated with the occurrence of the SSW
・WN1 EP-flux tends to propagate poleward and enhance upward
propagation

11. EOF1 of Predicted Zonal-Mean Zonal Wind Anomaly from Ensemble Mean on Dec. 12-14
Prediction from Dec. 5 and 6
EOF1: 36%
Contour: 0.5m/s;
Shaded region:
Significance Level > 95(99)%
5.4X104
2.0X107
・EOF1 shows the direction of the maximum spread among
ensemble
・EOF1 has a double jet structure, similar to the correlated
zonal wind profile with the occurrence of the SSW
・Thus, high sensitivity to the initial condition is observed
during the onset period of the SSW

12. Predicted Zonal Mean Temp. at 10hPa, 80N (K) α=-2,-1,0,1,2
Forecast Experiment by MRI/JMA AGCM from α×(regressed pattern)+(ensemble mean) on Dec. 13
Predicted Zonal Mean Temp. at 10hPa, 80N
(K)
α=-2,-1,0,1,2
Dec.
Observation

13. Predicted 3-day averaged Z10 for 27-29 Dec.
Forecasts from better condition
Observation α＝ α＝
180 90
270 α＝ α＝ α＝

14. Nonlinear Response to α Multiplied to the Anomaly
T(10hPa,80N) on Dec. 28 (K)
X: from Dec. 13
○: from Dec. 5 & 6
●: Obs.
Regressed Coef. of U on Dec.13 (α)
・The response is well described by a stepwise function of α
・Existence of a threshold value of α for the SSW
・Importance of blocking is confirmed by simplifying initial anomaly

15. Stratospheric Influence on Tropospheric Forecast Skill
Baldwin et al. (2003)
Statistical prediction of monthly-mean AO index using
a linear regression between NAM and AO index
Winter
The 150-hPa NAM predicts the monthly-mean AO better than the AO itself ← Downward migration of the NAM variation
Dynamical predictability of the NAM index winter (2001/ /06) JMA 1-month ensemble forecasts =130 forecasts
Dependence of forecast skill of the NAM index on phase of the NAM index in the stratosphere.

16. NAM index & Ensemble-Mean Forecast Error of NAM
03/04 Winter
04/05 Winter
Observed NAM index
Observed NAM index
Ensemble-Mean Error of NAM
Ensemble-Mean Error of NAM 10 20 30 10 20 30
lead time (days)
Prediction skill of the NAM index for 03/04 winter is better than that for 04/05 winter

17. NAM index & Spread of the Predicted NAM index
03/04 Winter
04/05 Winter
Observed NAM index
Observed NAM index
Spread of Predicted NAM
Spread of Predicted NAM 10 20 30 10 20 30
lead time (days)
Spread of the Predicted NAM index in the troposphere for 03/04 winter is also smaller than that for 04/05 winter

18. Dependence of Ensemble-Mean Forecast Error of predicted NAM index on the initial NAM index at 50 hPa
95%
Error
Classified based on NAM index at 50hPa
at the initial time of forecast
lead time (days)
p: >1: forecasts
n: forecasts
0: others forecasts
500hPa
●　Predicted NAM index error for negative NAM 　　case is significantly smaller than positive 　　NAM case for lead time from 6 to 12 days in 　　the upper troposphere, which suggests the stratospheric influence on the forecast skill of the prediction of the troposphere
●　For extended-range period, the error of predicted NAM for positive case becomes significantly large in the lower troposphere
250hPa

19. ・Improvement of error for group n is not so significant
Dependence of Ensemble-Mean Forecast Error of predicted NAM index on the initial NAM index at 1000hPa
1000hPa
95%
Error
p: >1: forecasts
n: forecasts
0: others forecasts
lead time (days)
500hPa
・Improvement of error for group n is not so significant
・Upward influence on the stratosphere
error difference p-n with significance 95(90)%
250hPa
lead time (days)

20. Summary (1)
Practical predictability of stratosphere-troposphere dynamical coupling during SSW events using operational 1-month ensemble forecast datasets provided by the JMA
(a) Tropospheric Influence on prediction of the stratosphere
Stratospheric sudden warming (SSW) event in Dec. ’01 was predictable at least 2 weeks in advance.
High sensitivity to the initial condition during the onset period of the SSW was observed for the prediction of zonal-mean temperature in the polar stratospheric region.
Characteristic zonal wind anomaly in the troposphere associated with the persistence of blocking was statistically related to the occurrence of the SSW.
Importance of blocking for the occurrence of the SSW was dynamically confirmed by a series of AGCM experiments.

21. Summary (2)
(b) Stratospheric influence on prediction of the troposphere
Forecast skill of the upper tropospheric NAM index for a lead time from 6 to 12 days tends to be better when the negative NAM index is observed in the stratosphere Propagating property of planetary waves after SSW? We could expect a better forecast skill of the tropospheric NAM index after SSW events
Improve understanding of stratospheric influence on troposphere
Reforecast experiments for a long-term
Increase member and frequency of 1-month ensemble forecast during SSW events

22. Stratospheric Influence on the troposphere after SSW Difference of 7-day averaged Z300 from Dec. 29-Jan. 1
Prediction from Dec. 5 & 6
(10 members of succeeded SSW prediction)－ (10 members of failed SSW prediction)
For succeed members, the positive anomaly over the Alaska and the negative anomaly over the East Asia tend to increase

23. Regressed forecast error of 7-day mean Z300 from Dec. 29-Jan. 4 w.r.t. Zonal-Mean Temp (10hPa, 80N) on Dec. 28
Prediction from Dec. 5 & 6
Anomaly from ensemble mean
Ensemble Mean
・Error of Z300 over Japan tends to decrease for members
which succeeded in predicting SSW

24. Ensemble-Mean Forecast Error of NAM
1000hPa
95%
Classified based on NAM index at 150hPa
at the initial time of forecast
Error
p: > forecasts
n: < forecasts
0: others forecasts
lead time (days)
statistical significance of p-n
500hPa
250hPa
lead time (days)

25. statistical significance of p-n
Dependence of Ensemble-Mean Forecast Error of predicted NAM index on the initial NAM index
1000hPa
95%
Error
Classified based on NAM index at 50hPa
at the initial time of forecast
lead time (days)
p: >1: forecasts
n: forecasts
0: others forecasts
500hPa
statistical significance of p-n
250hPa
lead time (days)

26. Cummulative WN1 Fz at 100hPa (40N-90N) of Dec. 6 Forecasts
Run 1
Run 2 13
December
January
・Deceleration of stratospheric zonal winds due to WN1 propagating from the troposphere
・Difference between Run 1 and Run 2 becomes evident after Dec.13

27. Regressed U anomaly on Dec. 12-14,2001
Squared Refractive Index (Q1)for WN1 (×a2)
Regressed U anomaly on Dec ,2001
Prediction from Dec. 5/6
Ensemble Mean
+Anomaly
-Anomaly
blue: negative, red: >150
・Large difference is seen in upper troposphere, but the profiles
in the stratosphere is basically the same
・Q1 for +anomaly in the upper troposphere is much larger than
Q1 for –anomaly due to Uyy associated with jet in polar region
・Q1 anomaly would deflect WN1 propagation poleward

28. Hindcast Experiments with MRI GCM
Predicted Zonal Mean Temp. at 10hPa, 80N (K)
JMA Model T106L40 Top: 0.4hPa
From Dec. 5 and 6, 2001
MRI GCM TL95L40 Top: 0.4hPa
From Dec. 5 and 6, 2001

29. Regressed Zonal-Mean Zonal Wind on Dec
Regressed Zonal-Mean Zonal Wind on Dec wrt Zonal-Mean Temp (10hPa, 80N) on Dec. 28
Prediction from Dec. 5/6
Anomaly from Ensemble Mean
Ensemble Mean
JMA Model
MRI GCM

30. Regressed Z300 on Dec. 12-14 wrt Zonal-Mean Temp (10hPa, 80N) on Dec
Prediction from Dec. 5/6
Anomaly from
Ensemble Mean
Ensemble Mean
JMA Model
MRI GCM