the dynamics and predictability of sudden stratospheric
TRANSCRIPT
Domeisen | ECMWF | November 2019
The dynamics and predictabilityof sudden stratospheric warmings
and their surface impacts
Daniela DomeisenAtmospheric Predictability, ETH Zurich
with contributions fromBernat Jiménez-Esteve, Alexander Wollert, Hilla Gerstman-Afargan,
Amy Butler, Andrew Charlton-Perez, Peter Hitchcock, Christian Grams, Lukas Papritz
Domeisen | ECMWF | November 2019
[m2
K / s
/ kg
]
video: A. Wollert
THE SUDDEN STRATOSPHERIC WARMING EVENT ON FEBRUARY 12, 2018
Feb 8 Feb 10 Feb 12
Feb 14 Feb 16 Feb 18
Figures: potential vorticity at 10hPa by Alexander Wollert
Domeisen | ECMWF | November 2019
video: A. Wollert
PART I: WHY DO WE CARE ABOUT THE STRATOSPHERE?
Part I. The diversity of sudden stratospheric warming events
Part II. The diversity of the tropospheric impact
The central date of a sudden stratospheric warming (SSW) event is not predictable beyond ~2 weeks.
perc
enta
ge o
f ens
embl
e m
embe
rs [%
] tha
t are
abl
e to
pre
dict
the
SSW
eve
nt
Domeisen | ECMWF | November 2019
manuscript submitted to JGR: Atmospheres
Figure 5. The average across all events of the percentage of ensemble members as a func-
tion of lead time [days] that detect the event within ± 3 days of the observed event for (a) early
stratospheric warming events, (b) strong polar vortex events, (c) SSW events, (d) negative heat
flux events, and (e) final warming events. The black line shows the multi-model mean based
on 5 prediction systems (CMA, ECCC, ECMWF, JMA, and UKMO). Dotted lines show where
25% and 75% of ensemble members detect the event. ’⇥’ marks the high-top models in the leg-
end. Where a prediction system was not used for the analysis or where there were not enough
available ensemble members (at least 10 members were required for a given lead time range) is
marked by an ⇥ in the color of the prediction system. Patterned black bars give the “false alarm
rate” (events that were predicted but not detected at the given lead times).
–13–
manuscript submitted to JGR: Atmospheres
Figure 5. The average across all events of the percentage of ensemble members as a func-
tion of lead time [days] that detect the event within ± 3 days of the observed event for (a) early
stratospheric warming events, (b) strong polar vortex events, (c) SSW events, (d) negative heat
flux events, and (e) final warming events. The black line shows the multi-model mean based
on 5 prediction systems (CMA, ECCC, ECMWF, JMA, and UKMO). Dotted lines show where
25% and 75% of ensemble members detect the event. ’⇥’ marks the high-top models in the leg-
end. Where a prediction system was not used for the analysis or where there were not enough
available ensemble members (at least 10 members were required for a given lead time range) is
marked by an ⇥ in the color of the prediction system. Patterned black bars give the “false alarm
rate” (events that were predicted but not detected at the given lead times).
–13–
manuscript submitted to JGR: Atmospheres
Figure 5. The average across all events of the percentage of ensemble members as a func-
tion of lead time [days] that detect the event within ± 3 days of the observed event for (a) early
stratospheric warming events, (b) strong polar vortex events, (c) SSW events, (d) negative heat
flux events, and (e) final warming events. The black line shows the multi-model mean based
on 5 prediction systems (CMA, ECCC, ECMWF, JMA, and UKMO). Dotted lines show where
25% and 75% of ensemble members detect the event. ’⇥’ marks the high-top models in the leg-
end. Where a prediction system was not used for the analysis or where there were not enough
available ensemble members (at least 10 members were required for a given lead time range) is
marked by an ⇥ in the color of the prediction system. Patterned black bars give the “false alarm
rate” (events that were predicted but not detected at the given lead times).
–13–
false alarm rate
S2S prediction systems (Vitart et al 2017):multi-model mean
Figure: M. Taguchi & A. Butler. From: Domeisen et al., 2019, JGR special issue on S2S prediction
WHAT IS THE PREDICTABILITY LIMITFOR SUDDEN STRATOSPHERIC WARMINGS? Part I – Part II
11 SSW events (1996 – 2010)
x = high-top models:
In the stratosphere (10hPa): Sudden Stratospheric Warming (SSW) events can be characterized as…
Jan 24, 2009Jan 5, 2004
displacement eventwave-1
split eventwave-2
DIFFERENT TYPES OFSUDDEN STRATOSPHERIC WARMINGS
Domeisen | ECMWF | November 2019
Part I – Part II
THE PREDICTABILITY DIFFERSFOR DIFFERENT TYPES OF SSW EVENTS
Split SSW events tend to be less predictable than displacement events(note the small sample size!)
Domeisen | ECMWF | November 2019
manuscript submitted to JGR: Atmospheres
Figure 6. Same as Fig. 5 but for SSW events separated into (a) displacement and (b) split
events. The black line corresponds to the multi-model mean from Figure 5c, the blue / red lines
indicate the multi-model mean for the displayed events only. A student t-test of the di↵erences
between the detection of splits and displacements gives the following p-values for lead times from
left to right: [0.6948,0.0279,0.7550,0.357,0.0925,0.3740]. The false alarm rates shown by the black
patterned bars are for all SSW events, as in Fig. 5c.
removing the model climatology (leaving the year to be corrected out) and then adding365
back ERA-interim climatology. The bias-correction had the strongest influence on the366
detection of strong vortex and negative heat flux events at long-leads (not shown). In367
particular, after bias-correction, a smaller percentage of members across prediction sys-368
tems detected strong vortex events at long lead times (suggesting an overestimation of369
these events in the model climatology), and a greater percentage of detected negative370
heat flux events at long lead times (suggesting an underestimation of these events in model371
climatology, in agreement with results from the the Coupled Model Intercomparison Project372
Phase 5 (CMIP5) models (Shaw et al., 2014, Fig. 5)).373
Figure 5 shows the percentage of ensemble members for each prediction system that374
detects the observed event within ± 3 days of its actual date, for lead times averaged375
over 5-day periods prior to the event, which occurs on day 0. The bin length is chosen376
as a balance between having su�cient hindcasts in each bin for each event while resolv-377
ing the lead times before each event. The “false alarm rate” is the percentage of mem-378
bers that predict an event to occur within a 1-30 day lead time when no event was ob-379
served. The comparison of the hit rate with the false alarm rate in Fig. 5 provides a mea-380
sure of the predictive skill.381
Below, we describe the di↵erences in the predictability between the di↵erent types382
of polar vortex events. The results should be prefaced by a number of caveats: 1) not383
all prediction systems produce a hindcast in each time bin for each event; 2) the num-384
ber of ensemble members varies across prediction systems; 3) the number of events is gen-385
erally small, due to the short period covered by the hindcasts; 4) hindcast data from dif-386
ferent model versions of a given model are sometimes used; 5) the ± 3-day window is an387
arbitrary choice which could matter for the accuracy in the detection of the events shown388
here; 6) the false alarm rates are used as a baseline for skill but the prediction systems389
could over- or underestimate these events, even after bias-correction; and 7) the percent-390
age of ensemble members forecasting an event is only one metric for the assessment of391
–14–
multi-model mean
perc
enta
ge o
f ens
embl
e m
embe
rs [%
] tha
t are
abl
e to
pre
dict
the
SSW
eve
nt
Can the difference in predictability of these events tell us anything about the different dynamics of these events?
Part I – Part II
Figure: M. Taguchi & A. Butler. From: Domeisen et al., 2019, JGR special issue on S2S prediction
6 displacement SSWs
5 split SSWs
SSW = sudden stratospheric warming
Data: S2S prediction systems (Vitart et al 2017)
x = high-top models:
THE DYNAMICS: SSWS ARE CAUSED BY UPWARD PROPAGATING WAVES
following Charney & Drazin (1961)
Upward propagation of planetary Rossby waves is limited to a range of background wind speeds whose limits depend on the wave characteristics
Domeisen | ECMWF | November 2019
zonal wavenumber
background zonal wind
critical velocity
zonal phase speed of the wave
0 < u� c < Uc = f(1/k2)<latexit sha1_base64="bFN2MSFc92SiS63QmPBBgADHH3o=">AAACCXicbZDLSgMxFIbP1Futt1GXboJFqAvrTBF0oVB047KC0xbaccikmTY0cyHJCGXo1o2v4saFIm59A3e+jWk7C63+EPjyn3NIzu8nnEllWV9GYWFxaXmluFpaW9/Y3DK3d5oyTgWhDol5LNo+lpSziDqKKU7biaA49Dlt+cOrSb11T4VkcXSrRgl1Q9yPWMAIVtryTGShc9T1scjSMTpCRN8cj6ALFFTs4+Fd7dAzy1bVmgr9BTuHMuRqeOZntxeTNKSRIhxL2bGtRLkZFooRTselbippgskQ92lHY4RDKt1suskYHWinh4JY6BMpNHV/TmQ4lHIU+rozxGog52sT879aJ1XBmZuxKEkVjcjsoSDlSMVoEgvqMUGJ4iMNmAim/4rIAAtMlA6vpEOw51f+C81a1dZ8c1KuX+ZxFGEP9qECNpxCHa6hAQ4QeIAneIFX49F4Nt6M91lrwchnduGXjI9vvNeWhw==</latexit><latexit sha1_base64="bFN2MSFc92SiS63QmPBBgADHH3o=">AAACCXicbZDLSgMxFIbP1Futt1GXboJFqAvrTBF0oVB047KC0xbaccikmTY0cyHJCGXo1o2v4saFIm59A3e+jWk7C63+EPjyn3NIzu8nnEllWV9GYWFxaXmluFpaW9/Y3DK3d5oyTgWhDol5LNo+lpSziDqKKU7biaA49Dlt+cOrSb11T4VkcXSrRgl1Q9yPWMAIVtryTGShc9T1scjSMTpCRN8cj6ALFFTs4+Fd7dAzy1bVmgr9BTuHMuRqeOZntxeTNKSRIhxL2bGtRLkZFooRTselbippgskQ92lHY4RDKt1suskYHWinh4JY6BMpNHV/TmQ4lHIU+rozxGog52sT879aJ1XBmZuxKEkVjcjsoSDlSMVoEgvqMUGJ4iMNmAim/4rIAAtMlA6vpEOw51f+C81a1dZ8c1KuX+ZxFGEP9qECNpxCHa6hAQ4QeIAneIFX49F4Nt6M91lrwchnduGXjI9vvNeWhw==</latexit><latexit sha1_base64="bFN2MSFc92SiS63QmPBBgADHH3o=">AAACCXicbZDLSgMxFIbP1Futt1GXboJFqAvrTBF0oVB047KC0xbaccikmTY0cyHJCGXo1o2v4saFIm59A3e+jWk7C63+EPjyn3NIzu8nnEllWV9GYWFxaXmluFpaW9/Y3DK3d5oyTgWhDol5LNo+lpSziDqKKU7biaA49Dlt+cOrSb11T4VkcXSrRgl1Q9yPWMAIVtryTGShc9T1scjSMTpCRN8cj6ALFFTs4+Fd7dAzy1bVmgr9BTuHMuRqeOZntxeTNKSRIhxL2bGtRLkZFooRTselbippgskQ92lHY4RDKt1suskYHWinh4JY6BMpNHV/TmQ4lHIU+rozxGog52sT879aJ1XBmZuxKEkVjcjsoSDlSMVoEgvqMUGJ4iMNmAim/4rIAAtMlA6vpEOw51f+C81a1dZ8c1KuX+ZxFGEP9qECNpxCHa6hAQ4QeIAneIFX49F4Nt6M91lrwchnduGXjI9vvNeWhw==</latexit><latexit sha1_base64="bFN2MSFc92SiS63QmPBBgADHH3o=">AAACCXicbZDLSgMxFIbP1Futt1GXboJFqAvrTBF0oVB047KC0xbaccikmTY0cyHJCGXo1o2v4saFIm59A3e+jWk7C63+EPjyn3NIzu8nnEllWV9GYWFxaXmluFpaW9/Y3DK3d5oyTgWhDol5LNo+lpSziDqKKU7biaA49Dlt+cOrSb11T4VkcXSrRgl1Q9yPWMAIVtryTGShc9T1scjSMTpCRN8cj6ALFFTs4+Fd7dAzy1bVmgr9BTuHMuRqeOZntxeTNKSRIhxL2bGtRLkZFooRTselbippgskQ92lHY4RDKt1suskYHWinh4JY6BMpNHV/TmQ4lHIU+rozxGog52sT879aJ1XBmZuxKEkVjcjsoSDlSMVoEgvqMUGJ4iMNmAim/4rIAAtMlA6vpEOw51f+C81a1dZ8c1KuX+ZxFGEP9qECNpxCHa6hAQ4QeIAneIFX49F4Nt6M91lrwchnduGXjI9vvNeWhw==</latexit>
The majority of waves are stationary (c = 0), but c ≠ 0 can have important consequences:
For c > 0, the propagation window opens towards higher wind speeds u.
For c < 0, the propagation window closes for high wind speeds u.
Part I – Part II
�
��
��
���� �����
�� �����
��
���������
���� �� ����������
��� �� �� �� � �������� ����� ;���= ��������
500hPa
eastwardwestward
zona
l wav
e nu
mbe
r
�
��
��
���� �����
�� �����
��
�B �� � ��
���� � ����������
��� �� � �� � �������� ����� ;���= ��������
10hPa
eastwardwestward
stat
iona
ry
westward eastwardwestward eastward
~ 5.5.km above ground ~ 35 km above ground
Domeisen | ECMWF | November 2019
Figures: Domeisen, Martius, Jiménez-Esteve, 2018, GRL
troposphere stratosphere
phase speed c [m/s]
HOW DO THESE CONSTRAINTS MANIFEST IN THE ATMOSPHERE?
Peak moves towards eastward propagating waves with heightzo
nal w
ave
num
ber
stat
iona
ry (c
=0)
If eastward propagating waves are favored for upward propagation, are these waves prominent before SSW events?
Increased eastward phase speed for model simulation with SSW events
Figure: Domeisen & Plumb, 2012, GRL
Figure 5. Comparison of the growth rates s [s!1] as a function of latitude (equation (1)), proportional to the Eady growthrate for the most unstable mode.
Figure 4. Phase speed spectra computed from geopotential height for zonal wave-2 at 189 hPa and 60 "S for (a) the controlrun, (b) the trunc12 run, and (c) the trunc2 run.
DOMEISEN AND PLUMB: ORIGIN OF TRAVELING PLANETARY WAVES L20817L20817
4 of 5
Figure 5. Comparison of the growth rates s [s!1] as a function of latitude (equation (1)), proportional to the Eady growthrate for the most unstable mode.
Figure 4. Phase speed spectra computed from geopotential height for zonal wave-2 at 189 hPa and 60 "S for (a) the controlrun, (b) the trunc12 run, and (c) the trunc2 run.
DOMEISEN AND PLUMB: ORIGIN OF TRAVELING PLANETARY WAVES L20817L20817
4 of 5
model run WITHOUT sudden stratospheric warming events
model run WITH (split) sudden stratospheric warming events
wave-2
wave-2
189hPa / 60⚬ lat
CAN SIMPLE MODELS REPRODUCE THIS?
Model: GFDL dry dynamical core (Held-Suarez)
Domeisen | ECMWF | November 2019
Part I – Part II
ARE EASTWARD PHASE SPEEDS FAVORED BEFORE SSWS IN REANALYSIS?
Increased eastward phase speed before split SSW events
Split SSW events are likely helped along by increased zonal phase speeds
Confidential manuscript submitted to Geophysical Research Letters
-20 -15 -10 -5 0 5 10 15 20
wave-1
0
0.05
0.1
0.15median SSW (days -19 to -6) = -1.03 median clim = -0.68KStest: not signif.
phase speed (100hPa)
-20 -15 -10 -5 0 5 10 15 20
ae
eeve
0
0.05
0.1
0.15median SSW (days -19 to -6) = -0.68 median clim = 0.04KStest: 99%
-20 -15 -10 -5 0 5 10 15 20
eve
0
0.05
0.1
0.15median SSW (days -19 to -6) = 1.67 median clim = -0.68KStest: 99%
zonal phase speed [m/s]-20 -15 -10 -5 0 5 10 15 20
wave-2
0
0.05
0.1
0.15= 1.03median SSW (days -19 to -6)
median clim = 0.04KStest: 99%
0 100 200 300 400 500 600 700
×10-3
0
1
2
3
4
5
6
7median SSW (days -19 to -6) = 296.91
median clim = 231.35KStest: 99%
amplitude (100hPa)
0 100 200 300 400 500 600 700
×10-3
0
1
2
3
4
5
6
7median SSW (days -19 to -6) = 186.51
median clim = 184.39KStest: not signif.
0 100 200 300 400 500 600 700
×10-3
0
1
2
3
4
5
6
7median SSW (days -19 to -6) = 231.27
median clim = 231.35KStest: 99%
wave amplitude [gpm]0 100 200 300 400 500 600 700
×10-3
0
1
2
3
4
5
6
7median SSW (days -19 to -6) = 192.93
median clim = 184.39KStest: not signif.
wave-2
wave-1
Figure 3. Phase speed [ms�1] (left) and amplitude [gpm] (right) at 100hPa for daily average values for
days -19 to -6 before a SSW event (gray bars) and for the NDJFM climatology (black bars). Additional hori-
zontal bars indicate the climatology where the gray bars are taller than the black bars. The median values for
the respective distributions and significant di�erences between the distributions according to a two-sample
Kolmogorov-Smirnov test are indicated. Values of 90%, 95%, and 99% significance are tested. Significance
values below 90% are indicated as ’not significant’. All distributions are normalized for comparison.
170
171
172
173
174
175
–16–
Figure: Domeisen, Martius, Jiménez-Esteve, 2018, GRL
eastwardwestward
westward eastward
Domeisen | ECMWF | November 2019
100hPa
100hPa
climatological phase speed distribution
phase speed distribution before a SSW event
Part I – Part II
Phase speed does NOT play a major role before displacementSSW events
climatological phase speed distribution
phase speed distribution before a SSW event
THIS IS NOT THE CASE FOR DISPLACEMENT SSWS
Figure: Domeisen, Martius, Jiménez-Esteve, 2018, GRL
eastwardwestward
westward eastward
100hPa
100hPa
Part I – Part II
Displacement events instead exhibit a large increase in amplitude of wave-1.
wave-2 is not helping with displacement events
Domeisen | ECMWF | November 2019
Do split and displacement SSW events exhibit different predictability because they are driven by a different mechanism?
Sudden stratospheric warming event on January 24, 2009
Confidential manuscript submitted to Geophysical Research Letters
-25 -20 -15 -10 -5 0 5phase speed [m/s]
10 15 20 25
ampl
itude
[gpm
]100
200
300
400
500
600
700
Jan 4
Jan 7Jan 10Jan 13Jan 16
Jan 19
-25 -20 -15 -10 -5 0 5phase speed [m/s]
10 15 20 25
ampl
itude
[gpm
]
100
200
300
400
500
600
700
Jan 4Jan 7
Jan 10
Jan 13Jan 16
Jan 19
wave-1
wave-2
Figure 4. Daily zonal phase speed [ms�1] vs wave amplitude [gpm] for November - March 1958-2013
(gray circles) and for days -20 to -5 before the 2009 SSW event (red circles) for wave-1 (top) and wave-2
(bottom) at 100 hPa. Every circle represents a daily average value and every third day before the 2009 SSW
event is indicated.
206
207
208
209
–17–
CASE STUDY: THE 2009 SPLIT SSW EVENT
Domeisen | ECMWF | November 2019
The acceleration and subsequent deceleration of the phase speed hints at a resonant effect responsible for the 2009 SSW event.see also: Matthewman & Esler, 2011; Plumb 2010
Figure: Domeisen, Martius, Jiménez-Esteve, 2018, GRL
days -20 to -5 before 2009 SSW event
Nov – Mar 1958-2013
Daily zonal phase speed vs wave amplitude for
Part I – Part II
SUMMARY – PART I
Domeisen | ECMWF | November 2019
Stratospheric eventPredictability: days to weeks
1. Eastward zonal phase speeds increase wave propagation into strong winds and can thereby help to efficiently weaken the vortex before SSW events.
2. Strong changes in phase speed can hint at resonant behavior.
3. The very unpredictable 2009 event shows strong signs of resonance, which might be responsible for the low predictability.
Domeisen, Martius, Jiménez-Esteve, 2019;see also: Matthewman & Esler, 2011
Predictability likely depends on the mechanism causing the event.
Part I – Part II
PART II: THE SURFACE IMPACT OF THE STRATOSPHERE
Domeisen | ECMWF | November 2019
Stratospheric eventPredictability: days to weeks
Tropospheric impactPredictability: weeks to months
What is the surface impact of SSWs and what does it depend on?
WHY DO WE CARE ABOUT SUDDEN STRATOSPHERIC WARMING EVENTS?
Figure: modified from cpc.ncep.noaa.gov
positiveNAO
negative NAO
OBSERVATIONS
Negative NAO tendency after a sudden stratospheric warming event changes the weather over Europe
SSW event
Domeisen | ECMWF | November 2019
Part I – Part II
Domeisen | ECMWF | November 2019
warm
cold, snow
cold, dry
warm, rain
warm, rain
dry
cold
warmjet streamL
H
L
H
Sudden stratospheric warming events lead to a wavy jet stream with cold weather in northern and central Europe.
Negative NAO tendency after a sudden stratospheric warming event changes the weather over Europe
Positive NAO:early Dec 2017 – mid-February 2018
Negative NAO: After SSW event:Late February 2018 – end of March 2018
Part I – Part II
WHY DO WE CARE ABOUT SUDDEN STRATOSPHERIC WARMING EVENTS?
WHAT IS THE“DOWNWARD IMPACT” OF SSW EVENTS?
Domeisen | ECMWF | November 2019
all SSW (24)SSW not followed by persistent NAO or switch to negative NAO (8)
SSW followed by persistent NAO (14)
Figure: Domeisen, 2019, JGR-Atmospheres 500hPa geopotential height [m]
days
8 to
52
afte
r SSW
eve
nt
A negative NAO / NAM event is generally seen as the response to SSW events. see e.g. Baldwin & Dunkerton, 2001; Charlton-Perez et al, 2018; Karpechko et al 2017
NAO = North Atlantic OscillationNAM = Northern Annular Mode
Part I – Part II
Shading: values significant at p < 0.05 level.
NOT ALL SSW EVENTS EXHIBITA “DOWNWARD IMPACT”
Domeisen | ECMWF | November 2019
all SSW (24)
500hPa geopotential height [m]Figure: Domeisen, 2019, JGR-Atmospheres
SSW followed by persistent NAO (14)
This signal is dominated by the ~ two thirds of SSW events that exhibit a negative NAO signal. Domeisen, 2019.see also: Charlton-Perez et al, 2018; Karpechko et al 2017
days
8 to
52
afte
r SSW
eve
nt
NAO = North Atlantic Oscillation
Part I – Part II
Shading: values significant at p < 0.05 level.
NOT ALL SSW EVENTS EXHIBITA “DOWNWARD IMPACT”
Domeisen | ECMWF | November 2019
all SSW (24)
500hPa geopotential height [m]Figure: Domeisen, 2019, JGR-Atmospheres
SSW not followed by persistent NAO or switch to negative NAO (8)
SSW followed by persistent NAO (14)
days
8 to
52
afte
r SSW
eve
nt
Note: in turn, only 25% of persistent negative NAO events are preceded by a SSW event.
Part I – Part II
Shading: values significant at p < 0.05 level.
THE DIVERSITY IN THE SURFACE IMPACT OF SSWS MANIFESTS IN THE STORM TRACK RESPONSE
Domeisen | ECMWF | November 2019
man
uscr
ipt
subm
itte
dto
Geo
phys
ical
Res
earc
hLe
tter
s
Fig
ure
1.C
ompo
site
sof
the
trop
osph
eric
circ
ulat
ion
aver
aged
over
the
30da
ysfo
llow
ing
(top
row
)al
l24
hist
oric
alSS
Wev
ents
,and
SSW
even
tsas
soci
ated
wit
h(m
iddl
e)eq
uato
rwar
dan
d
(bot
tom
)po
lew
ard
Atl
anti
cje
tsh
ifts
(16
and
8ev
ents
,res
pect
ivel
y)in
the
ER
A-I
nter
imre
anal
y-
sis
(197
9–20
14).
Col
orsh
adin
gre
pres
ents
the
(a,d
,g)
MSL
Pan
omal
ies
(hPa)
,(b,
e,h)
zona
lwin
d
anom
alie
s(m
s-1)
at30
0hP
a,an
d(c
,f,i)
stor
mtr
ack
inte
nsity
,exp
ress
edus
ing
EK
E(M
Jm
-2),
vert
ical
lyin
tegr
ated
from
1000
to20
0hP
a.A
nom
alie
sar
eco
mpu
ted
wit
hre
spec
tto
the
daily
cli-
mat
olog
y.B
lack
cont
ours
repr
esen
tth
eD
JFcl
imat
olog
yof
(b,e
,h)
300-
hPa
zona
lwin
d(c
onto
ur
inte
rval
10m
s-1,s
tart
ing
from
10m
s-1),
and
(c,f,
i)E
KE
(con
tour
inte
rval
0.1
MJ
m-2
,sta
rtin
g
from
0.3
MJ
m-2
).A
nom
alie
ssi
gnifi
cant
atth
e95
%le
vel(
usin
ga
Stud
ent’
st-
test
)ar
een
clos
ed
bygr
ayco
ntou
rs.
The
red
box
in(b
)in
dica
tes
the
area
wit
hth
est
rong
est
SSW
influ
ence
on
zona
lwin
din
the
mid
lati
tude
Nor
thA
tlan
tic.
The
brac
kets
inth
eti
tle
ofea
chpa
neli
ndic
ate
the
num
ber
ofSS
Ws
inea
chco
mpo
site
.
355
356
357
358
359
360
361
362
363
364
365
366
–19–
equatorward Atlantic jet (16)
poleward Atlantic jet (8)
zona
l win
d an
omal
y at
30
0hPa
eddy
kin
etic
ene
rgy
anom
aly
inte
grat
ed
from
100
0 to
200
hPa
from: Afargan-Gerstman & Domeisen, in rev.
all SSW events (24)
Figure: tropospheric circulation averaged over 30 days following SSW events
response to SSWs:equatorward shifted jet in Atlantic
response to SSWs:weaker storm track in Atlantic
Part I – Part II
Domeisen | ECMWF | November 2019
man
uscr
ipt
subm
itte
dto
Geo
phys
ical
Res
earc
hLe
tter
s
Fig
ure
1.C
ompo
site
sof
the
trop
osph
eric
circ
ulat
ion
aver
aged
over
the
30da
ysfo
llow
ing
(top
row
)al
l24
hist
oric
alSS
Wev
ents
,and
SSW
even
tsas
soci
ated
wit
h(m
iddl
e)eq
uato
rwar
dan
d
(bot
tom
)po
lew
ard
Atl
anti
cje
tsh
ifts
(16
and
8ev
ents
,res
pect
ivel
y)in
the
ER
A-I
nter
imre
anal
y-
sis
(197
9–20
14).
Col
orsh
adin
gre
pres
ents
the
(a,d
,g)
MSL
Pan
omal
ies
(hPa)
,(b,
e,h)
zona
lwin
d
anom
alie
s(m
s-1)
at30
0hP
a,an
d(c
,f,i)
stor
mtr
ack
inte
nsity
,exp
ress
edus
ing
EK
E(M
Jm
-2),
vert
ical
lyin
tegr
ated
from
1000
to20
0hP
a.A
nom
alie
sar
eco
mpu
ted
wit
hre
spec
tto
the
daily
cli-
mat
olog
y.B
lack
cont
ours
repr
esen
tth
eD
JFcl
imat
olog
yof
(b,e
,h)
300-
hPa
zona
lwin
d(c
onto
ur
inte
rval
10m
s-1,s
tart
ing
from
10m
s-1),
and
(c,f,
i)E
KE
(con
tour
inte
rval
0.1
MJ
m-2
,sta
rtin
g
from
0.3
MJ
m-2
).A
nom
alie
ssi
gnifi
cant
atth
e95
%le
vel(
usin
ga
Stud
ent’
st-
test
)ar
een
clos
ed
bygr
ayco
ntou
rs.
The
red
box
in(b
)in
dica
tes
the
area
wit
hth
est
rong
est
SSW
influ
ence
on
zona
lwin
din
the
mid
lati
tude
Nor
thA
tlan
tic.
The
brac
kets
inth
eti
tle
ofea
chpa
neli
ndic
ate
the
num
ber
ofSS
Ws
inea
chco
mpo
site
.
355
356
357
358
359
360
361
362
363
364
365
366
–19–
equatorward Atlantic jet (16)
poleward Atlantic jet (8)
zona
l win
d an
omal
y at
30
0hPa
eddy
kin
etic
ene
rgy
anom
aly
inte
grat
ed
from
100
0 to
200
hPa
from: Afargan-Gerstman & Domeisen, in rev.
all SSW events (24)
Figure: tropospheric circulation averaged over 30 days following SSW events
negative NAO response
negative NAO response
Part I – Part II
THE DIVERSITY IN THE SURFACE IMPACT OF SSWS MANIFESTS IN THE STORM TRACK RESPONSE
IS THE SURFACE IMPACT IN THE NORTH ATLANTIC MODULATED BY THE PACIFIC?
Domeisen | ECMWF | November 2019
man
uscr
ipt
subm
itte
dto
Geo
phys
ical
Res
earc
hLe
tter
s
Fig
ure
1.C
ompo
site
sof
the
trop
osph
eric
circ
ulat
ion
aver
aged
over
the
30da
ysfo
llow
ing
(top
row
)al
l24
hist
oric
alSS
Wev
ents
,and
SSW
even
tsas
soci
ated
wit
h(m
iddl
e)eq
uato
rwar
dan
d
(bot
tom
)po
lew
ard
Atl
anti
cje
tsh
ifts
(16
and
8ev
ents
,res
pect
ivel
y)in
the
ER
A-I
nter
imre
anal
y-
sis
(197
9–20
14).
Col
orsh
adin
gre
pres
ents
the
(a,d
,g)
MSL
Pan
omal
ies
(hPa)
,(b,
e,h)
zona
lwin
d
anom
alie
s(m
s-1)
at30
0hP
a,an
d(c
,f,i)
stor
mtr
ack
inte
nsity
,exp
ress
edus
ing
EK
E(M
Jm
-2),
vert
ical
lyin
tegr
ated
from
1000
to20
0hP
a.A
nom
alie
sar
eco
mpu
ted
wit
hre
spec
tto
the
daily
cli-
mat
olog
y.B
lack
cont
ours
repr
esen
tth
eD
JFcl
imat
olog
yof
(b,e
,h)
300-
hPa
zona
lwin
d(c
onto
ur
inte
rval
10m
s-1,s
tart
ing
from
10m
s-1),
and
(c,f,
i)E
KE
(con
tour
inte
rval
0.1
MJ
m-2
,sta
rtin
g
from
0.3
MJ
m-2
).A
nom
alie
ssi
gnifi
cant
atth
e95
%le
vel(
usin
ga
Stud
ent’
st-
test
)ar
een
clos
ed
bygr
ayco
ntou
rs.
The
red
box
in(b
)in
dica
tes
the
area
wit
hth
est
rong
est
SSW
influ
ence
on
zona
lwin
din
the
mid
lati
tude
Nor
thA
tlan
tic.
The
brac
kets
inth
eti
tle
ofea
chpa
neli
ndic
ate
the
num
ber
ofSS
Ws
inea
chco
mpo
site
.
355
356
357
358
359
360
361
362
363
364
365
366
–19–
all SSW events (24) equatorward Atlantic jet (16)
poleward Atlantic jet (8)
zona
l win
d an
omal
y at
30
0hPa
eddy
kin
etic
ene
rgy
anom
aly
inte
grat
ed
from
100
0 to
200
hPaThe opposite
responses in the North Atlantic storm track also exhibit opposite “precursors” in the eastern North Pacific!
The troposphere can have a strong impact on the manifestation of the downward response to SSWs. see also: Garfinkel et al 2013, Chan & Plumb, 2009
Figure: tropospheric circulation averaged over 30 days following SSW events
from: Afargan-Gerstman & Domeisen, in rev.
Part I – Part II
TO WHAT EXTENT DOES THE SURFACE IMPACT OF SSWSDEPEND ON THE STATE OF THE TROPOSPHERE?
Domeisen | ECMWF | November 2019
(a) (b)
(c) (d)
[hPa]
[hPa]
[hPa]
[hPa]
[std]
Figure 3. Standardized geopotential height anomalies for the sector -80�E to 40�E / 60�N to 90�N for (a) all SSW events, and (b - d) sub-
divided by the weather regime that is dominant at the onset of the SSW event as indicated in the panel titles. Hatching (stippling) indicates
that the confidence intervals and the random distributions overlap by less than 25% (10%).
is again illustrated by a case study of the SSW events on Feb 12, 2018 (Fig. 4c). This anomaly is not highly robust, but it is
nevertheless significantly different from a random sample at the 25 % level. Several SSWs with a cyclonic regime at the onset
are followed by GL at a longer lag (Fig. 2d), thus, likely causing these anomalies. Still the GL frequencies only reach 25% at
most and also other regimes occur more often albeit with low frequencies around 25%. These findings and the small amplitude
of the anomalies suggest that the variability in the tropospheric response to SSWs is large after a cyclonic regime at lag zero -220
which is also confirmed by the inspection of individual cases (not shown).
5 Impact on Surface Weather
Since each weather regime is associated with characteristic surface weather, the modulation of regime successions in the
aftermath of a SSW by the tropospheric state at the time of a SSW is likely a key contributor to the marked variability in
the surface impact, such as, for example, 2m temperature. Hence, we here consider spatial composites of 2m temperature225
anomalies and anomalies of 500 hPa geopotential height (Z500’) for the three groups of SSW events discussed in the previous
sections (Fig. 5).
During SSWs with GL at the onset, initially strong warm anomalies prevail over Greenland and the Canadian Archipelago,
whereas western Russia and Scandinavia are anomalously cold (Fig. 5a). Consistent with the subsequent progression of weather
regimes - typically towards the cyclonic AT regime or EuBL - mild conditions are established throughout central Europe (from a230
9
Domeisen, Grams, Papritz, in prep.
The events with “downward impact” are dominated by events that show blocking over Europe at the onset of the SSW.
These regimes show a favor for transitioning into Greenland blocking, the “canonical response” to SSWs.
at onset of SSW (day 0) at onset of SSW (day 0)
at onset of SSW (day 0)
Units: Standard deviation of geopotential height anomalies for Atlantic sector.
(a) (b)
(c) (d)
[hPa]
[hPa]
[hPa]
[hPa]
[std]
Figure 3. Standardized geopotential height anomalies for the sector -80�E to 40�E / 60�N to 90�N for (a) all SSW events, and (b - d) sub-
divided by the weather regime that is dominant at the onset of the SSW event as indicated in the panel titles. Hatching (stippling) indicates
that the confidence intervals and the random distributions overlap by less than 25% (10%).
is again illustrated by a case study of the SSW events on Feb 12, 2018 (Fig. 4c). This anomaly is not highly robust, but it is
nevertheless significantly different from a random sample at the 25 % level. Several SSWs with a cyclonic regime at the onset
are followed by GL at a longer lag (Fig. 2d), thus, likely causing these anomalies. Still the GL frequencies only reach 25% at
most and also other regimes occur more often albeit with low frequencies around 25%. These findings and the small amplitude
of the anomalies suggest that the variability in the tropospheric response to SSWs is large after a cyclonic regime at lag zero -220
which is also confirmed by the inspection of individual cases (not shown).
5 Impact on Surface Weather
Since each weather regime is associated with characteristic surface weather, the modulation of regime successions in the
aftermath of a SSW by the tropospheric state at the time of a SSW is likely a key contributor to the marked variability in
the surface impact, such as, for example, 2m temperature. Hence, we here consider spatial composites of 2m temperature225
anomalies and anomalies of 500 hPa geopotential height (Z500’) for the three groups of SSW events discussed in the previous
sections (Fig. 5).
During SSWs with GL at the onset, initially strong warm anomalies prevail over Greenland and the Canadian Archipelago,
whereas western Russia and Scandinavia are anomalously cold (Fig. 5a). Consistent with the subsequent progression of weather
regimes - typically towards the cyclonic AT regime or EuBL - mild conditions are established throughout central Europe (from a230
9
Hatching (stippling): confidence intervals and the random distributions overlap by less than 25% (10%).
Cyclonic regimes: zonal regime, Atlantic trough, Scandinavian trough
Part I – Part II
Negative NAO response to sudden stratospheric warming event in reanalysis
500hPa geopotential height anomaly for days 15 – 60 after an SSW event [m]
Domeisen | ECMWF | November 2019
ERAinterim(32 SSW events)
HOW ABOUT THE ROLE OF THE STRATOSPHERE FOR THE SURFACE IMPACT?
Domeisen, Hitchcock et al, in prep.
MPI-ESM model(894 SSW events)
Part I – Part II
Data: MPI-ESM seasonal prediction model: Baehr et al 2015
Negative NAO response in reanalysis and seasonal forecasting model is considerably stronger for persistent SSW events (Polar Jet Oscillation events)
500hPa geopotential height anomaly for days 15 – 60 after an SSW event [m]
Domeisen | ECMWF | November 2019
HOW ABOUT THE ROLE OF THE STRATOSPHERE FOR THE SURFACE IMPACT?
Domeisen, Hitchcock et al, in prep.
ERAinterim(13 persistent SSW events)
MPI-ESM model(403 persistent SSW events)
Part I – Part II
Data: MPI-ESM seasonal prediction model: Baehr et al 2015
SUMMARY PART II: THE SURFACE IMPACT OF THE STRATOSPHERE
Stratospheric eventPredictability: days to weeks
Swiss Data Science Center
Only two thirds of SSW events are followed by the “canonical” negative NAO event. Domeisen, 2019, Charlton-Perez et al, 2018, Karpechko et al 2017
The storm track weakens for these two thirds of SSW events. The eastern Pacific might influence the storm track response over the North Atlantic.Afargan-Gerstman & Domeisen, in rev.
The “downward response” of SSW events depends on the tropospheric state at the onset of the SSW and the persistence of the signal in the lower stratosphere.Domeisen, Grams, Papritz, in prep..see also: Garfinkel et al 2013, Chan & Plumb, 2009
Eastern Pacific precursor Tropospheric impact
Predictability: weeks to months
Thank you!@Domeisen_DSSW competition: