1
2
3
4
Weather regime transitions and the interannual
variability of the North Atlantic Oscillation. Part
II: Dynamical processes
5
6
Dehai Luo and Jing Cha
7
RCE-TEA, Institute of Atmospheric Physics, Chinese Academy of Sciences,
8
Beijing, China
9
Steven B, Feldstein
10
Department of Meteorology, Pennsylvania State University, University Park,
11
Pennsylvania
12
Submitted to J. Atmos. Sci. for a revised version
13
Corresponding author address: Dr. Dehai Luo, RCE-TEA, Institute of Atmospheric
14
Physics, Chinese Academy of Sciences, Beijing, China, Email: ldh@mail.iap.ac.cn
15
16
17
Abstract
1
2
In this study, our attention is focused on identifying the dynamical processes that
3
contribute to the NAO  NAOand NAO   NAOtransitions that occur within
4
1978-90 (P1) and 1991-2008 (P2). By constructing Atlantic ridge (AR) and
5
Scandinavian blocking (SBL) indices, the composite analysis demonstrates that in a
6
stronger AR (SBL) winter a NAO  ( NAO  ) event can more easily transit into a
7
NAO  (NAO  ) event. Composites of 300-hPa geopotential height anomalies for
8
the NAO   NAO  and NAO   NAO  transition events within P1 and P2 are
9
calculated. It is shown for P2 (P1) that the NAO  to SBL to NAO  (NAO  to
10
AR to NAO  ) transition results from the retrograde drift of an enhanced high
11
latitude large-scale positive (negative) anomaly over northern Europe during the
12
decay of the previous NAO  (NAO  ) event. This finding cannot be detected for
13
NAO events without transition.
14
Moreover, it is found that the amplification of retrograding wavenumber 1 is
15
more important for the NAO   NAO  transition within P1, but the marked
16
re-intensification and retrograde movement of both wavenumbers 1 and 2 after the
17
NAO  event decays is crucial for the NAO   NAO  transition within P2. It is
18
further shown that destructive (constructive) interference between wavenumbers 1
19
and 2 over the North Atlantic within P1 (P2) is responsible for the subsequent weak
20
NAO  (strong NAO  ) anomaly associated with the NAO   NAO  (NAO  
21
NAO  ) transition. Moreover, the weakening (strengthening) of the vertically1
1
integrated zonal wind (upstream Atlantic storm track) is found to play an important
2
role in the NAO regime transition.
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
2
1
1. Introduction
2
In Part I of this study (Luo et al. 2012a), we have established a plausible
3
connection between intraseasonal NAO regime transitions and interannual
4
variability of the winter mean NAO index. In that study, it was shown that the
5
frequencies of occurrence of the NAO  NAO and NAO   NAO transition events
6
differ between 1978-1990 (P1) and 1991-2008 (P2). Such a difference between P1
7
and P2 results in a significant change in the interannual variability of the NAO
8
pattern. It is further suggested that there is a likely connection between the NAO
9
transitions and the Atlantic ridge (AR) and Scandinavian blocking (SBL)
10
teleconnection patterns. However, it is unclear as to how the NAO transitions
11
depend on the occurrence of AR and SBL events. More precisely, it is unclear what
12
dynamical processes contribute to the NAO  NAO (NAO   NAO) transition
13
within P1(P2) and what are their characteristics. The main purpose of the present
14
study is to examine these problems.
15
Since NAO regime transitions have been found to take place on intraseasonal
16
time scales (Luo et al. 2012a), it is inferred that intraseasonal changes in the NAO
17
must be related to the activity of planetary-scale waves also on intraseasonal time
18
scales. Sawyer (1970) first noted that the blocking occurrence over the Atlantic
19
sector resulted from retrograding planetary waves in high latitudes. Branstator
20
(1987) and Kushnir (1987) have revealed that retrograding large-scale disturbances
21
with a period of 3-4 weeks are frequently active in high latitudes between 50 N
3
1
and 70 N , and are often associated with blocking events over the North Atlantic
2
and North Pacific. Michelangeli and Vautard (1998) found that a high-latitude
3
retrograding zonal wavenumber 1 pattern contributes significantly toward the onset
4
of the Euro-Atlantic blocking. Recent investigations have further revealed that the
5
two phases of the NAO correspond to two different flow regimes: a high-latitude
6
(Greenland) blocking and a zonal (non- blocking) flow (Luo et al. 2007; Woollings
7
et al. 2010). Hannachi (2010) noted that the sectorial weather regimes reflect
8
mainly blocking and non-blocking flows. Thus, it is concluded that the NAO
9
regime transition is probably connected to retrograding high-latitude low-frequency
10
disturbances.
11
This study is organized as follows: In section 2, we first present composites of
12
daily NAO indices in terms of the winter mean AR and SBL strengths. For these
13
calculations the AR and SBL indices are defined in a manner similar to that in
14
Woollings et al. (2011). Composites of the AR and SBL indices are then separately
15
performed for the NAO  to NAO  and NAO  to NAO  transition events. It is
16
found that the NAO  to NAO  (NAO  to NAO  ) transition is more likely to
17
occur when the AR (SBL) is enhanced, with the route NAO  to AR to NAO 
18
( NAO  to SBL to NAO  ) taking place. In section 3, we perform composites of
19
the 300-hPa geopotential height anomalies for the NAO  to NAO  and NAO  to
20
NAO  transition events within P1 and P2, respectively. It is shown that the
21
re-intensification and retrograde shift of high latitude low-frequency disturbances
4
1
over northern Europe is extremely important for the NAO regime transitions. In
2
Section 4, a spectral decomposition of the composite 300-hPa geopotential height
3
field is made for the transition events. It is found that the two NAO transition
4
events exhibit different wavenumber characteristics with destructive (constructive)
5
intereference between wavenumbers 1 and 2 being important for the weakening
6
(strengthening) of the NAO  (NAO  ) pattern. In section 5, additional dynamical
7
factors affecting the NAO regime transitions are presented. The main conclusion
8
and a discussion are given in section 6.
9
2. Variations of the Atlantic ridge and Scandinavian blocking patterns and
10
their relationship with NAO regime transitions
11
In this paper, the data set used is the National Centers for Environmental
12
Prediction-National Center for Atmospheric Research (NCEP-NCAR) daily mean,
13
multi-level, gridded ( reanalyses from December 1950 to February 2009. The
14
definitions of the NAO events for both phases and associated transition events can
15
be found in Part I (Luo et al. 2012a). Briefly, the normalized daily NAO index
16
being less (greater) than or equal to  1.0 (  1.0 ) standard deviation, with at least 3
17
consecutive days, is defined as a NAO  (NAO  ) event. A NAO transition event is
18
defined to include both NAO  and NAO  events, whose total life period from the
19
beginning of a NAO  (NAO  ) event to the end of a NAO  (NAO  ) is less than or
20
equal to 45 days. In the present paper, the AR (SBL) index is further defined to
21
reflect the variation of the AR (SBL) pattern as can be seen from Fig.2 of Luo et al.
5
1
(2012a).
2
In this section, we first indicate that the NAO   NAO  (NAO   NAO  )
3
transitions are linked to the occurrence of AR (SBL) events. Before doing this
4
calculation, 300-hPa geopotential height composites for NAO events without
5
transition are performed (see Fig.1). According to the composite field in this figure,
6
we can define the AR (SBL) pattern. As Fig.1 indicates, for the positive phase there
7
are two positive anomalies: one over the subtropical Atlantic ( 20  50 N ) and the
8
other over northeastern Europe ( 50  90 N ), which are most evident at Lag 0 day.
9
The two anomalies are reversed for the negative phase (Fig.1b). It is also evident in
10
Fig. 1a that although the two positive anomalies can exhibit a change in intensity
11
from Lag -6 to Lag +4, they do not present a sign variation. The two anomalies
12
correspond in fact to the positive anomalies of the AR and SBL patterns,
13
respectively. As noted below, the AR and SBL patterns will undergo marked
14
changes in strength and sign once a NAO transition event takes place. The cluster
15
analysis results of Luo et al.(2012a) indicate that the NAO regime transitions are
16
more likely to be related to changes in the strengths of AR and SBL patterns.
17
However, this relationship between the NAO transitions and the variations of the
18
AR and SBL patterns in intensity and sign was not examined in detail in that work.
19
Here, we present the AR and SBL indices for the NAO regime transitions. One
20
important reason for introducing the AR and SBL indices is that these two indices
21
reflect how the NAO transitions depend upon the strength and sign variations of the
6
1
AR and SBL patterns. Another is that we use the two indices to test our viewpoint
2
that the occurrence of NAO transition events can be attributed mainly to the
3
activity of enhanced retrograding high latitude low-frequency anomalies.
4
Similar to Woollings et al. (2011) the strength of the AR (SBL) is defined as the
5
regional mean of the 300-hPa daily geopotential height anomaly over the sector
6
30  45 N , 40W  5 E ( 50  80 N , 10  50 E ) for all winter seasons. The strengths
7
of the AR and SBL patterns at each day are defined by the daily AR and SBL
8
indices used here.
9
From the daily AR and SBL indices defined above, the winter (DJF) mean AR
10
and SBL indices are determined. The normalized winter mean AR and SBL indices
11
are shown in Fig.2 for the period of 1978-2008, respectively. In this figure, index
12
values greater (less) than or equal to 0.2 (-0.2) standard deviations are defined as
13
corresponding to strong (weak) AR and SBL patterns. Following this definition,
14
NAO  (NAO  ) events are selected for strong and weak AR (SBL) winters during
15
1978-2008. Then, composites of the daily NAO index are determined for all NAO 
16
(NAO  ) events that coincide with strong and weak AR (SBL) winters (see Fig.3).
17
It is seen from the composite of daily NAO  indices that the composite NAO 
18
index can decay to a positive value for the AR being strong (Fig.3a). Thus, NAO 
19
events can more easily transit into NAO  events when the AR is relatively strong.
20
In contrast, it is extremely difficult for transition events to occur when the AR is
21
relatively weak. An analogous result is found for NAO  to NAO  transition
7
1
events when the SBL is strong (Fig.3b). The difference of the composite NAO
2
index between the weak and strong AR (SBL) strengths is statistically significant at
3
the 90% confidence level with a Monte-Carlo test. This result indicates that an
4
enhanced AR (SBL) pattern favors the transition from the NAO  (NAO  ) to the
5
NAO  (NAO  ) event. This provides an explanation for the result noted by Luo et
6
al. (2012a) using k-means clustering.
7
Although the result in Fig. 3 shows that NAO   NAO  (NAO   NAO  )
8
transition events can more readily take place for stronger AR (SBL) winters, the
9
concise relationship between the AR (SBL) variation and NAO transition events is
10
unclear. In this case, it is necessary to perform composites of the daily AR and SBL
11
indices for NAO non-transition and transition events. Figure 4 shows separate
12
composites of the daily AR and SBL indices for non-transition and transition events
13
following the definition of NAO transition events presented by Luo et al. (2012a).
14
It is found that for the case with the NAO  to NAO  transition the AR index is
15
negative between Lag -9 and Lag +6 and becomes positive after Lag +6 (solid line
16
in Fig.4a). At the same time, the SBL index (solid line) is positive between Lag -6
17
and Lag 0 and then becomes negative. Thus, the strengthening (weakening) of the
18
AR (SBL) is followed both by the transition from a NAO  to a NAO  event and
19
an enhanced positive AR (negative SBL) index during the subsequent NAO 
20
process. In some sense, because of the projection of the AR onto the NAO  (Fig. 2
21
in Part I of this study), the NAO  to NAO  transition coincides with changes in
8
1
the strength of the AR pattern. At the beginning of the NAO  to NAO  transition,
2
when the NAO index is negative, a negative (positive) height anomaly over Europe
3
is evident in higher (lower) latitudes. As the NAO  event transits into a NAO 
4
event, the enhanced high-latitude (low-latitude) negative (positive) anomaly over
5
Europe retrogrades, which then leads to a strengthening of the AR pattern. Thus,
6
the peak of the AR index in Fig. 4a (solid line) occurs during the period from Lag
7
+11 to Lag +17, consistent with the composite NAO index of the NAO  to NAO 
8
transition events (not shown). This behavior can also be seen from the composite
9
geopotential height anomalies for NAO transition events, as shown in the next
10
section. Thus, it appears that the NAO   NAO  transition is accomplished
11
through the route NAO  to AR to NAO  .
12
Prior to the NAO  event transiting into a NAO  event, the SBL is generally
13
stronger because of the NAO being positive (Luo et al. 2007). It is found from Fig.
14
4b that during the NAO  to NAO  transition the SBL strength is positive between
15
Lag-10 and Lag+12, indicating a tendency to increase before Lag 0 and then
16
decrease until Lag +4. A rapid re-intensification of the SBL pattern is again seen
17
after Lag+4, followed by the beginning of its decay at about Lag+7, after which it
18
becomes negative at Lag +12. Thus, the transition from a NAO  to a NAO  event
19
is closely related to a change in the SBL strength. Moreover, it is seen that the
20
occurrence of the SBL precedes the formation of a NAO  event during the NAO 
21
transition (solid line in Fig. 4b). This is because the NAO  event arises from the
9
1
retrograde displacement of the SBL pattern. Feldstein (2003) found that European
2
blocks retrograde before a NAO  event forms. More recently, Sung et al. (2011)
3
suggested that NAO  events tend to be preceded by a blocking ridge in the vicinity
4
of northern Europe. Cassou et al. (2008) noted that an enhanced SBL, and NAO  ,
5
associated with the Madden-Julian Oscillation (MJO), can be interpreted as being
6
the consequence of the previous NAO  excitation. However, we see here that the
7
re-intensification of the SBL pattern is crucial for the NAO   NAO  transition.
8
As noted in the next section, such a SBL re-intensification is attributed to the
9
enhanced retrograding high latitude positive anomaly over northern Europe. At the
10
same time, it is also found that the AR tends to weaken because of the growing
11
NAO  anomaly. This behavior is extremely difficult to see for the case without the
12
NAO regime transition. As a result, the NAO   NAO  transition is, to some extent,
13
accomplished by the route NAO  to SBL to NAO  through the re-intensification
14
and westward drift of the SBL pattern. Thus, the NAO   NAO  (NAO   NAO  )
15
transition is closely related to the strengthening of the AR (SBL) via the route
16
NAO  to AR to NAO  (NAO  to SBL to NAO  ).
17
3. NAO regime transition and retrograding high-latitude low-frequency
18
disturbances
19
a) Role of retrograding high-latitude low-frequency disturbances in the NAO
20
regime transition
10
1
To understand why the NAO regime transitions are related to the occurrence of
2
enhanced AR and SBL patterns, we perform composites of 300-hPa geopotential
3
height anomalies for the NAO  to NAO  (NAO  to NAO  ) transition events in
4
P1 (P2) (see Fig. 5).
5
For the composite NAO  to NAO  transition events during P1, at Lag -6, a
6
positive anomaly is present that extends from the Greenland to northern Europe,
7
and a weak negative anomaly is seen in lower latitudes. This dipole anomaly is
8
amplified into a typical NAO  pattern (Fig. 5a at Lag 0). A large-scale negative
9
anomaly is seen over northern Europe once the NAO  anomaly begins to decay
10
(Fig. 5a from Lag +2 to Lag +4). At this time, we can see that this high-latitude
11
negative anomaly begins to intensify and undergoes a retrograde shift. When the
12
retrograding large-scale negative anomaly and the positive anomaly to its south
13
side enter the Atlantic basin, a typical NAO  pattern is established (Fig. 5a at Lag
14
+16) and continues until Lag +20. Thus, it appears that during the NAO  to NAO 
15
transition process, the retrograding high-latitude large-scale negative anomaly
16
becomes more active and prominent. During this time period, the AR (SBL)
17
strength, as shown in Fig. 4a, is enhanced (weakened) due to the strengthening and
18
retrograding high latitude negative anomaly that moves into the Atlantic basin.
19
Such a high latitude large-scale anomaly appears to correspond to the retrograding
20
high latitude low-frequency disturbance found by Branstator (1987), Kushnir
21
(1987), and Kushnir and Wallace (1989). Thus, the enhancement of the
11
1
retrograding large-scale negative anomaly (low-frequency disturbance) over
2
northern Europe is likely to result in the NAO  to NAO  transition via the
3
strengthening of the AR pattern. It is also found that the NAO  dipole anomaly in
4
Fig. 5a (after Lag +12) is relatively weak most likely because the Atlantic storm
5
track eddies are weaker during P1 than in P2 (Luo et al. 2011). Moreover, it is seen
6
that the AR pattern confined in the region 40  W  5  E is enhanced from Lag +6 to
7
Lag +16 in Fig. 5a although the center of the positive anomaly in the subtropical
8
Atlantic for the same lags is located more eastward ( 0   30  E ). This is because this
9
positive anomaly can strengthen the AR anomaly as it moves westward. The AR
10
anomaly is negative before the transition and becomes positive during the transition
11
toward the NAO+ event. This result can be evidently seen from the variation of the
12
composite AR index for the NAO  to NAO  transition (solid line in Fig. 4a). Of
13
course, the intensification of the AR pattern can also be reflected by the variation of
14
the high-latitude negative height anomaly originating from the northern Europe.
15
The high-latitude negative anomaly corresponds in fact to a large-scale
16
Scandinavian trough (ST), which is opposite to the SBL. When the ST is intensified
17
and undergoes a westward shift, the AR pattern is enhanced and a NAO  to NAO 
18
transition event can occur (Fig.5a from Lag + 6 to Lag +16). This implies that the
19
ST or negative SBL index should lead the AR index, as shown in Fig.4a. Thus, it is
20
concluded that the NAO  to NAO  transition is completed by the path NAO  to
21
AR to NAO  or NAO  to ST to NAO  because the AR and negative SBL
12
1
indices exhibit a consistent variation, even though the AR index lags slightly the
2
negative SBL index.
3
For the composite NAO  to NAO  transition events, a weak NAO  pattern is
4
seen over the Atlantic basin (Fig. 5b at Lag -5). This weak NAO  anomaly evolves
5
into a typical NAO  pattern (Fig. 5b at Lag 0). During the NAO  growth process, a
6
large-scale positive anomaly or a SBL pattern, is seen over northern Europe during
7
the period from Lag-5 to Lag 0. This high-latitude positive anomaly tends to decay
8
once the NAO  anomaly also decays. However, this anomaly re-intensifies after
9
Lag +3 and undergoes retrograde movement (Fig. 5b from Lag +5 to Lag +15).
10
When the amplified positive anomaly enters the Atlantic basin, it combines with the
11
relatively weak, and also retrograding, negative anomaly to its south, to finally
12
form a typical NAO  pattern (Fig. 5b from Lag +9 to Lag +15). This NAO 
13
pattern persists until Lag +19. This demonstrates that an enhanced retrograding
14
high-latitude positive anomaly over Europe plays an important role in the NAO 
15
to NAO  transition. Since the enhanced high latitude positive anomaly from Lag
16
+3 to Lag +7 is over the European continent, one can understand this
17
re-intensification of the SBL pattern (Fig. 4b) as corresponding with the
18
retrograding high-latitude positive anomaly over Europe. When the enhanced high
19
latitude positive anomaly enters the Atlantic basin and migrates farther toward the
20
west, the SBL pattern will decay. As a result, the SBL pattern, defined as a positive
21
anomaly over northern Europe, is weakened during the period from Lag +9 to Lag
13
1
+15 even though the retrograding high latitude positive anomaly is markedly
2
strengthened in the Atlantic basin. Thus, it is concluded that the NAO  to NAO 
3
transition is accomplished by the route NAO  to SBL to NAO  via the
4
re-intensification and retrograde shift of the SBL pattern (Fig. 5b at Lag+7).
5
Sawyer (1970) and Lejena¨s and Madden (1992) found that the occurrence of
6
European blocking is linked with retrograding high-latitude low-frequency
7
disturbances. Michelangeli and Vautard (1998) noted that the Euro-Atlantic
8
blocking can result from a retrograding zonal wavenumber 1 anomaly. However, in
9
this study, we have proposed that the SBL or European blocking acts as a bridge
10
connecting the NAO  to the NAO  when the anomalies move into the Atlantic
11
basin. In other words, when the SBL is re-intensified and undergoes marked
12
retrograde movement, the NAO  to NAO  transition becomes more likely. More
13
recently, Michel and Riviere (2011) have connected the weather regime transitions
14
to the type of Rossby wave breaking and its occurrence frequency. Nevertheless,
15
we can conclude from discussions here that the NAO  to NAO  (NAO  to
16
NAO  ) transition events are, to large extent, attributed to the re-intensification of
17
the retrograding high-latitude large-scale positive (negative) anomaly over Europe
18
via the enhanced SBL (AR) pattern during P2 (P1).
19
b) Impact of the NAO regime transition on the zonal position of the subsequent
20
NAO pattern
21
It is interesting to examine if the NAO regime transition can affect the zonal
14
1
position of the subsequent NAO pattern. We can see from a comparison between
2
Fig.1 and Fig. 5 that during the NAO  to NAO  transition the NAO  pattern is
3
located farther eastward (Lag +10 to Lag +16 in Fig. 5a) relative to the case
4
without transition events, as seen in Fig. 1a (Lag-2 to Lag +2). This conclusion also
5
holds for the NAO
6
displacement of the NAO  pattern induced by the NAO  to NAO  transition
7
appears to be more distinct. In fact, to some extent these results can be explained in
8
part by the phase speed formula for a finite amplitude Rossby wave or the
9
destructive (constructive) interference between zonal wavenumbers 1 and 2 within

to NAO

transition events. However, the eastward
10
P1 (P2), as found in the next section.
11
4. Spectral evolution of the NAO transition events and destructive
12
(constructive) interference between planetary waves
13
a) Wavenumber structure of the composite NAO pattern
14
To obtain the spectral evolution of the composite NAO anomaly, zonal Fourier
15
decomposition (Bloomfield 1976; Colucci et al. 1981) is used to detect the time
16
variation of each wave component of the NAO  to NAO  transition events within
17
P2 and the NAO  to NAO  transition events within P1. For these two cases, the
18
time evolution of each wave component for zonal wavenumbers 1-3 is shown in
19
Fig. 6.
20
For the NAO  to NAO  transition event during P1, it is found that the amplitude
15
1
of wavenumber 1 is large throughout the NAO  event and during the early stage of
2
the transition into a NAO  event. In contrast, the amplification of wavenumbers 2
3
and 3 is relatively weak. The amplitude of wavenumber 1 reaches its maximum
4
value at Lag +2. After that time, the amplitudes of all three wave components
5
remain small. In contrast, for P2, wavenumbers 1 and 2 reach their maximum
6
amplitudes at Lag 0. Afterwards, the two wave components decay and then undergo
7
a re-intensification during the period from Lag +6 to Lag +18.
8
anticipated that this re-intensification plays an important role in the NAO  to
9
NAO  transition during P2.
10
It is to be
b) Destructive (constructive) interference between wavenumbers 1 and 2
11
It should be pointed out that the phase relationship between wavenumbers 1 and
12
2 is also important for the transition between the NAO  and NAO  . This can be
13
seen by looking at the separate spatial evolution of each wave component
14
associated with the NAO  to NAO  and NAO  to NAO  transitions (not shown).
15
Destructive (constructive) interference between wavenumbers 1 and 2 within P1
16
(P2) can be seen with Hovmöller diagrams of the composite high-latitude height
17
anomaly (Fig. 7). We can see within P1 that the westward movement of the high
18
latitude positive anomaly of wavenumber 1 in the Atlantic region ( 60W  0 ) is
19
evidently more rapid than that of wavenumber 2 before Lag +9 (Fig. 7a). Within P2
20
the westward propagation speeds of the high latitude positive anomalies of
21
wavenumbers 1 and 2 over the European continent ( 0  60 E ) are almost the same
16
1
after Lag 0 (Fig. 7b). Thus, it is concluded that there is destructive (constructive)
2
interference of the high-latitude anomalies between wavenumbers 1 and 2 over the
3
North Atlantic within P1 (P2). This destructive (constructive) interference tends to
4
result in the weakening (strengthening) of the subsequent NAO  (NAO  ) anomaly.
5
The destructive (constructive) interference between wavenumbers 1 and 2 within
6
P1 (P2) can be explained in terms of the phase speed of a finite amplitude Rossby
7
wave obtained from a weakly nonlinear framework (Luo 2000; Luo et al. 2011).
8
The phase speed of the finite amplitude Rossby wave can be approximately
9
expressed as CP  u0 
(   Fu0 )
M 0

2
2
k m F
2k
2
(where u0 is a uniform mean westerly
10
wind, k and m are the zonal and meridional wavenumbers of the NAO anomaly
11
with amplitude M 0 respectively,   0 and F  1 is generally chosen) (Luo 2000).
12
As a result, it is evident that within P1 wavenumbers 1 and 2 have different phase
13
speeds because they are of different amplitude and undergo markedly different
14
retrograde movement, leading to destructive interference between the two waves
15
over the North Atlantic due to their phase speed difference. Thus, it is possible that
16
a relatively weak NAO  anomaly can arise from the NAO  to NAO  transition
17
because the anomaly of zonal wavenumber 2 tends to counteract the anomaly of
18
zonal wavenumber 1 due to their destructive interference.
19
The eastward displacement of the induced NAO  (NAO  ) pattern due to the
20
NAO  to NAO  (NAO  to NAO  ) transition can be seen from a comparison
21
between Fig. 1 and Fig. 5. During the NAO  to NAO  (NAO  to NAO  )
17
1
transition process the subsequent NAO  (NAO  ) anomaly within P1 (P2) will be
2
located more eastward because it stems from the retrograde shift of enhanced
3
high-latitude positive (negative) anomaly over the northern Europe and cannot
4
reach the farther upstream region of the Atlantic basin during the life period of the
5
NAO  (NAO  ) event. Although the zonal position of the observed NAO anomaly
6
is dominated by many factors (Ulbrich and Christoph 1999), the NAO  to NAO 
7
(NAO  to NAO  ) transition can be thought of as being a new mechanism for
8
affecting the strength and zonal position of the NAO anomaly.
9
c) Zonal movement of high latitude anomalies over the European continent for
10
NAO transition and non-transition events
11
To illustrate the difference between the high-latitude anomaly movement for
12
NAO transition and non-transition events, we show in Fig. 8 the temporal evolution
13
of the zonal position of the maximum amplitude of the high-latitude negative
14
(positive) anomaly over northern Europe for the composite NAO  to NAO 
15
(NAO
16
longitudinal position of the maximum amplitude of the corresponding high latitude
17
composite anomalies for NAO events without transition. It is evident for the NAO
18
events without transition that the high-latitude negative (positive) anomaly over
19
northern Europe is almost stationary, and thus cannot move into the Atlantic basin.
20
However, for NAO transition events, the high-latitude negative (positive) anomaly
21
moves rapidly westward and enters the central Atlantic when the NAO event

to NAO

) transition events. Correspondingly, Fig. 9 shows the
18
1
undergoes a transition from the NAO  (NAO  ) to NAO  (NAO  ) phase. Thus,
2
the re-intensification and westward movement of the retrograding high-latitude
3
negative (positive) anomaly over the European continent appears to be especially
4
crucial for the NAO  to NAO  (NAO  to NAO  ) transition.
5
5. Dynamical factors contributing to NAO regime transitions
6
More recently, Franzke et al. (2011) emphasized a dynamical link between
7
preferred regime transitions and shifts of the Atlantic jet. Michel and Riviere (2011)
8
noted that the weather regime transition is linked to Rossby wave breaking.
9
However, Luo et al. (2011) found that when the Atlantic storm track intensity is
10
sufficiently strong, the preferred regime transition from the NAO  to NAO  is
11
more likely although it is probably modulated by the MJO in the tropics (Cassou
12
2008; Lin et al. 2009). In this section, we will reveal some additional dynamical
13
features that contribute to the NAO regime transitions. As indicated by various
14
numerical studies (e.g., Ulbrich and Christoph 1999), the Atlantic mean westerly
15
wind and storm track strengths are two important factors that affect the phase of the
16
NAO event. Thus, examining the difference of the Atlantic mean westerly wind and
17
storm track strengths prior to the NAO occurrence between transition and non-
18
transition events can be very helpful for improving our understanding of NAO
19
regime transitions. Here, the composite vertically-averaged (a vertical mean
20
between the 300- and 850-hPa levels) zonal wind and storm track (defined as the
21
2.5-7.0 day eddy kinetic energy (EKE)) averaged during the period from Lag-15 to
19
1
Lag-10 are considered as the barotropic zonal wind and storm track prior to the
2
NAO onset, respectively. These quantities are referred to as the prior barotropic
3
zonal wind and prior storm track, respectively.
4
The difference between the prior (lag -15 to Lag -10) barotropic zonal wind for
5
the transition and non-transition events is plotted in Fig. 10. It is seen for both P1
6
and P2 that prior to the NAO onset the anomalous barotropic zonal winds has
7
dominant negative values in mid-high latitudes and positive values in much lower
8
and higher latitudes. Because the regions of the weakened zonal winds correspond
9
to the NAO region, it is thought that the weakening of the prior barotropic zonal
10
winds in mid-high latitudes over the North Atlantic basin or upstream of the NAO
11
region is more distinct for transition events than for non-transition events.
12
Correspondingly, Fig. 11 shows the difference of the 300-hPa EKE in the Atlantic
13
basin averaged during the period from Lag-15 to Lag-10 between transition and
14
non-transition events. It is found for both P1 and P2 that the prior Atlantic storm
15
track is stronger in the upstream region of the NAO pattern for transition events
16
than for non-transition events. This finding is supported by the observational and
17
theoretical results of Luo et al. (2011). This hints that within P1 and P2 the marked
18
intensification of the prior upstream Atlantic storm track is an important contributor
19
to the NAO regime transition. A similar conclusion about changes in the winter
20
mean barotropic zonal wind and Atlantic storm track can also be seen from a
21
comparison between NAO transition and non-transition winters (not shown). Thus,
20
1
it is conjectured from the results obtained in this section that the weakening
2
(strengthening) of the prior barotropic zonal wind (upstream Atlantic storm track) is
3
an important dynamical factor that contributes to the NAO regime transition. This
4
is because the high latitude positive (negative) anomaly will be enhanced and
5
undergo a marked westward shift when the prior barotropic zonal wind in mid-high
6
latitudes (upstream Atlantic storm track) is weakened (intensified). In this case, the
7
intensification and westward displacement of the high latitude positive (negative)
8
anomaly will replace the previous anomaly over the Atlantic basin which leads to
9
the transition from the NAO  (NAO  ) to NAO  (NAO  ) event.
10
It is possible that the NAO sign transitions are related to the MJO (Cassou et al.
11
2008; Lin et al. 2009). The role of the MJO in affecting the transition between
12
NAO  and NAO  events deserves further examination, and will be explored in
13
future work.
14
6 Conclusion and discussion
15
In this paper, we have examined the dynamical processes that contribute to the
16
NAO  to NAO  (NAO  to NAO  ) transition within P1 (P2). It was shown that
17
the NAO  to NAO  (NAO  to NAO  ) transition is closely related to changes in
18
the Atlantic ridge (AR) and Scandinavian blocking (SBL) patterns. During stronger
19
AR (SBL) winters, the NAO  to NAO  (NAO  to NAO  ) transition is more
20
likely to take place. This transition follows the route NAO  to AR to NAO 
21
(NAO  to SBL to NAO  ) within P1 (P2). By further calculating composites of the
21
1
300-hPa geopotential height anomalies, it was revealed that the NAO  to SBL to
2
NAO  (NAO  to AR to NAO  ) transition events within P2 (P1) can be attributed
3
to the re-intensification and retrograde movement of a large-scale positive (negative)
4
anomaly over northern Europe. Michelangeli and Vautard (1998) have discussed
5
the contribution of the retrograding wavenumber 1 to Atlantic blocking. Michel and
6
Rivière (2011) have recently analyzed the NAO  to SBL to NAO  transition and
7
have shown that the SBL to NAO  transition is mainly due to cyclonic wave
8
breaking. In that sense, the retrograde displacement of the positive height anomaly
9
can be viewed as triggered by nonlinearities. Cassou (2008) also noted that the
10
NAO  to SBL to NAO  transition tends to be excited by the MJO in tropics.
11
However, Luo et al. (2011) found that an excessively strong storm track can lead to
12
the NAO  to SBL to NAO  transition in that the European blocking is more
13
easily enhanced and undergoes a marked westward drift. This result does not
14
contradict the results of Ulbrich and Christoph (1999), who found that the positive
15
trend of the NAO from 1970s is due to the presence of a stronger storm-track. This
16
is because the NAO  anomaly is enhanced as the Atlantic storm track is
17
strengthened, as long as the storm track is not too strong. In contrast, it can play the
18
reversed role and contribute to the NAO  to NAO  transition if the Atlantic storm
19
track is sufficiently strong. This result has been shown by Luo et al. (2011) from
20
both observational and theoretical perspectives. In the present study, we present a
21
new finding that enhanced retrograding high latitude low-frequency negative
22
1
(positive) anomalies over the European continent play an important role in the
2
NAO  to NAO  (NAO  to NAO  ) transition.
3
Within P2, the NAO  to NAO  transition is dominated by wavenumbers 1 and 2
4
that undergo re-intensification and retrograde movement after the NAO  anomaly
5
decays. In contrast, within P1, the NAO  to NAO  transition is mainly dominated
6
by wavenumber 1. It is also seen that there is destructive (constructive) interference
7
between wavenumbers 1 and 2 for the NAO  to NAO  (NAO  to NAO  )
8
transition within P1 (P2). This can be understood in terms of the longitudinal
9
westward phase speed of wavenumber 1 being much larger than (almost the same
10
as) that of wavenumber 2 within P1 (P2). Such destructive (constructive)
11
interference over the North Atlantic affects the strength and zonal position of the
12
subsequent NAO anomaly. Dynamical factors that contributing to the NAO regime
13
transition are also examined in this paper. It is found that the weakening
14
(strengthening) of the prior barotropic zonal wind (upstream Atlantic storm track)
15
in the NAO region makes an important contribution to the NAO regime transition
16
because downstream high latitude large-scale anomalies are enhanced and move
17
westward.
18
Although our present study provides observational evidence for the role played
19
by the weakening of the prior barotropic zonal wind and the strengthening of the
20
prior upstream Atlantic storm track in the NAO regime transition, a theoretical
21
study of this process was not made. This is an unsolved problem that we plan to
23
1
address in a future study.
2
3
Acknowledgments
4
5
The authors acknowledge the support from the “one Hundred Talent Plan” of the
6
Chinese Academy of Sciences (Y163011) and National Science Foundation of
7
China (41075042, 40921004) and National Science Foundation grants ATM-
8
0852379 and AGS-1036858. The authors would like to thank two anonymous
9
reviewers for valuable comments that improved this paper.
10
11
12
13
14
15
16
17
18
19
20
21
24
1
References
2
3
4
5
6
Bloomfield, P., 1976: Fourier Analysis of Time series: An introduction. Wiley,
258pp.
Branstator, G., 1987: A striking example of the atmosphere’s leading traveling
pattern. J. Atmos. Sci., 44, 2310-2323.
7
Cassou, C., 2008: Intraseasonal interaction between the Madden–Julian Oscillation
8
and the North Atlantic Oscillation. Nature, 455, 523-527, doi:10.1038 /nature
9
07286.
10
Colucci, S. J., A. Z. Loesch and L. F. Bosart, 1981: Spectral evolution of a blocking
11
episode and comparison with wave interaction theory. J. Atmos. Sci., 38, 2092-
12
2111.
13
Franzke, C., T. Woollings and O. Martius, 2011: Persistent circulation regimes and
14
preferred regime transitions in the North Atlantic, J. Atmos. Sci., 68, 2809–
15
2825.
16
17
18
19
Hannchi, A., 2010: On the origin of planetary-scale extratropical winter circulation
regimes. J. Atmos. Sci., 67, 1382-1401.
Kushnir, Y., 1987: Retrograding wintertime low-frequency disturbances over the
North Pacific Ocean. J. Atmos. Sci., 44, 2727-2742.
25
1
Kushnir, Y. and J. M. Wallace, 1989: Low-frequency variability in the Northern
2
Hemisphere winter: Geographical distribution, structure and time-scale
3
dependence. J. Atmos. Sci., 46, 3122-3142.
4
5
Lejena¨s, H. and R. A. Madden, 1992: Traveling planetary-scale waves and
blocking. Mon. Wea. Rev., 120, 2821-230
6
7
Lin, H., G. Brunet and J. Derome, 2009: An observed connection between the North
8
Atlantic Oscillation and the Madden-Julian Oscillation. J. Climate, 22, 364-
9
380.
10
Luo, D., 2000: Planetary-scale baroclinic envelope Rossby solitons in a two-layer
11
model and their interaction with synoptic-scale eddies. Dyn. Atmos. Oceans,
12
32, 27-74.
13
Luo D., A. Lupo and H. Wan, 2007: Dynamics of eddy-driven low frequency dipole
14
modes. Part I: A simple model of North Atlantic Oscillations. J. Atmos. Sci.,
15
64, 3-38.
16
Luo, D., Z. Zhu, R. Ren, L. Zhong and C. Wang, 2010: Spatial pattern and zonal
17
shift of the North Atlantic Oscillation. Part I: A dynamical interpretation. J.
18
Atmos. Sci., 67, 2805-2826.
19
Luo, D., Y. Diao and B. S. Feldstein, 2011: The variability of the Atlantic storm
20
track and the North Atlantic Oscillation: A link between intraseasonal and
21
interannual variability. J. Atmos. Sci., 68, 577-601.
22
Luo, D., J. Cha and B. S. Feldstein, 2012a: Weather regime transitions and the
26
1
interannual variability of the North Atlantic Oscillation: Part I: A likely
2
connection. J. Atmos. Sci. (in press).
3
4
5
6
7
8
9
Michel,C. and G. Rivi`ere, 2011: The link between Rossby wave breakings and
weather regime transitions, J. Atmos. Sci., 68, 1730-1745.
Michelangeli, P. A. and R. Vautard, 1998: The dynamics of Euro-Atlantic blocking
onsets. Quart. J. Roy. Meteor. Soc., 124, 1045-1070.
Sawyer, J. S., 1970: Observational characteristics of atmospheric fluctuations with
a time scale of a month. Quart. J. Roy. Meteor. Soc., 96, 610-625.
Sung, M. K., G. H. Lim, J. S. Kug and S. I. An, 2011: A linkage between the North
10
Atlantic Oscillation and its downstream development due to the existence of a
11
blocking ridge . J. Geophys. Res., 116, D11107,doi:10.1029/2010JD01506.
12
Ulbrich, U., and M. Christoph, 1999: A shift of the NAO and increasing storm track
13
activity over Europe due to anthropogenic greenhouse gas forcing. Climate
14
Dyn., 15, 551-559.
15
Woollings, T. J., A. Hannachi, B. J. Hoskins, and A. G. Turner, 2010: A regime
16
view of the North Atlantic Oscillation and its response to anthropogenic forcing.
17
J. Climate, 23,1291-1307.
18
Woollings, T. J., J. G. Pinto and J. A. Santos, 2011: Dynamical evolution of North
19
Atlantic ridges and poleward jet stream displacements. J. Atmos. Sci., 68,
20
954-963.
21
27
1
Figure captions:
2
3
Figure 1. Time sequences of composited geopotential height anomalies at 300 hPa
4
for non-transition NAO events: (a) NAO  anomaly (the contour interval (CI) is 40
5
gpm) and (b) NAO  anomaly (CI=40 gpm). The dark (light) shading indicates
6
positive (negative) value regions that exceed the 95% confidence level for a
7
two-sided student’s t-test.
8
9
10
Figure 2. Time series of normalized winter (DJF) mean AR and SBL indices, in
which the dashed line represents  0.2 standard deviations.
11
12
Figure 3. Composites of daily NAO indices for strong and weak winter mean AR
13
and SBL, in which the solid line represents strong AR and SBL, and the dashed line
14
denotes weak AR and SBL: (a) composite of NAO  events for the different AR
15
strength and (b) composite of NAO  events for the different SBL strength.
16
17
Figure 4. Composites of the daily Atlantic ridge (AR) and Scandinavian blocking
18
(SBL) indices for NAO  (NAO  ) events without transition and for NAO  to
19
NAO  (NAO  to NAO  ) transition events for P1 (P2), in which the dashed (solid)
20
line represents the NAO events without transition (transition events): (a) P1 and (b)
21
P2.
28
1
2
Figure 5. Time sequences of composited geopotential height anomalies at 300 hPa
3
for NAO  to NAO  transition events within P1 and for NAO  to NAO 
4
transition events within P2, in which the dashed (solid) line denotes the negative
5
(positive) value: (a) NAO  to NAO  transition (the contour interval (CI) is 40 gpm)
6
and (b) NAO  to NAO  transition (CI=40 gpm). The dark (light) shading
7
indicates positive (negative) value regions that exceed the 95% confidence level for
8
a two-sided student’s t-test.
9
10
Figure 6. Time sequences of the amplitude of each wave component for
11
wavenumber 1-3 in the composited geopotential height anomalies for NAO
12
transition events, in which Lag 0 denotes the maximum amplitude of the NAO
13
anomaly before the transition: (a) NAO  to NAO  transition in P1 and (b) NAO 
14
to NAO  transition in P2.
15
16
Figure 7. Hovmöller diagrams of the composite geopotential height anomaly
17
averaged from 60 N
18
NAO  to NAO  transition and (b) NAO  to NAO  transition. The thick bar
19
denotes the propagation of the maximum.

to 75 N
as a function of longitude and time (Lag days): (a)
20
21
Figure 8．Time variation of the zonal position of the maximum amplitude of the
29
1
large-scale high-latitude negative (positive) anomaly over the European continent
2
and its downstream region for the composite NAO anomaly from the NAO  to
3
NAO  (NAO  to NAO  ) transition, in which “-30” and “30” denote “ 30W ” and
4
“ 30 E ” respectively, and the solid (dashed) curve represents wavenumber 1
5
(wavenumber 2) : (a) NAO  to NAO  transition events and (b) NAO  to NAO 
6
transition events.
7
8
Figure 9. As Fig.8, except for the composite NAO anomaly for NAO events
9
without transition: (a) NAO  and (b) NAO  .
10
11
Figure 10. Difference of the composited barotropic zonal wind averaged from
12
Lag-15 to Lag-10 between the transition and non-transition events within P1 (a)
13
and P2 (b), in which the shading indicates the region above the 80% confidence
14
level for a two-sided student’s t-test.
15
16
Figure 11. As Fig.10, except for the eddy kinetic energy at 300-hPa.
17
18
19
20
21
22
30
1
2
3
4
5
6
(a)
7
8
9
31
1
2
3
4
5
6
(b)
7
Figure 1. Time sequences of composited geopotential height anomalies at 300 hPa
8
for non-transition NAO events: (a) NAO  anomaly (the contour interval (CI) is 40
9
gpm) and (b) NAO  anomaly (CI=40 gpm). The dark (light) shading indicates
10
positive (negative) value regions that exceed the 95% confidence level for a
11
two-sided student’s t-test.
32
1
2
3
4
5
6
Figure 2. Time series of normalized winter (DJF) mean AR and SBL indices, in
7
which the dashed line represents  0.2 standard deviations.
8
9
10
11
12
33
1
2
3
(a)
4
5
6
(b)
7
Figure 3. Composites of daily NAO indices for strong and weak winter mean AR
8
and SBL, in which the solid line represents strong AR and SBL, and the dashed line
9
denotes weak AR and SBL: (a) composite of NAO  events for the different AR
10
strength and (b) composite of NAO  events for the different SBL strength.
11
34
1
2
3
4
(a)
5
6
7
(b)
8
9
Figure 4. Composites of the daily Atlantic ridge (AR) and Scandinavian blocking
10
(SBL) indices for NAO  (NAO  ) events without transition and for NAO  to
11
NAO  (NAO  to NAO  ) transition events for P1 (P2), in which the dashed (solid)
12
line represents the NAO events without transition (transition events): (a) P1 and (b)
13
P2.
14
15
16
35
1
2
3
4
5
6
7
8
(a)
36
1
2
3
4
5
6
7
8
37
1
(b)
2
Figure 5. Time sequences of composited geopotential height anomalies at 300 hPa
3
for NAO  to NAO  transition events within P1 and for NAO  to NAO 
4
transition events within P2, in which the dashed (solid) line denotes the negative
5
(positive) value: (a) NAO  to NAO  transition (the contour interval (CI) is 40 gpm)
6
and (b) NAO  to NAO  transition (CI=40 gpm). The dark (light) shading
7
indicates positive (negative) value regions that exceed the 95% confidence level for
8
a two-sided student’s t-test.
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
38
1
2
3
(a)
4
5
6
7
8
9
10
11
12
13
39
1
2
3
(b)
4
Figure 6. Time sequences of the amplitude of each wave component for
5
wavenumber 1-3 in the composited geopotential height anomalies for transition
6
events, in which Lag 0 denotes the maximum amplitude of the NAO anomaly
7
before the transition: (a) NAO  to NAO  transition in P1 and (b) NAO  to NAO 
8
transition in P2.
9
10
11
12
13
40
1
(a)
2
3
4
(b)
5
Figure 7. Hovmöller diagrams of the composite geopotential height anomaly
6
averaged from 60 N
7
NAO  to NAO  transition and (b) NAO  to NAO  transition. The thick bar
8
denotes the propagation of the maximum.

to 75 N
as a function of longitude and time (Lag days): (a)
9
41
1
(a)
2
(b)
3
Figure 8, Time variation of the zonal position of the maximum amplitude of the
4
large-scale high-latitude negative (positive) anomaly over the European continent
5
and its downstream region for the composite NAO anomaly from the NAO  to
6
NAO  (NAO  to NAO  ) transition, in which “-30” and “30” denote “ 30W ” and
7
“ 30E ” respectively, and the solid (dashed) curve represents wavenumber 1
8
(wavenumber 2) : (a) NAO  to NAO  transition events and (b) NAO  to NAO 
9
transition events.
10
11
12
13
14
15
16
17
42
1
2
(a)
(b)
3
Figure 9, As Fig.8, except for the composite NAO anomaly for NAO events
4
without transition: (a) NAO  and (b) NAO  .
5
6
7
8
9
10
11
12
13
14
15
43
1
(a)
2
3
(b)
4
5
Figure 10. Difference of the composited barotropic zonal wind averaged from
6
Lag-15 to Lag-10 between the transition and non-transition events within P1 (a)
7
and P2 (b), in which the shading denotes the region above the 80% confidence level
8
for a two- sided Student’s t-test.
9
10
11
12
13
44
1
2
(a)
3
4
5
(b)
Figure 11. As Fig.10, except for the eddy kinetic energy at 300-hPa.
45