ENSO SIGNATURE IN NORTH EUROPEAN TIME SERIES OF ICE CONDITIONS DETECTED BY SINGULAR SPECTRUM ANALYSI

1
ENSO SIGNATURE IN NORTH EUROPEAN TIME SERIES OF
ICE CONDITIONS DETECTED BY SINGULAR SPECTRUM
ANALYSIS AND WAVELET TRANSFORM

• Aslak Grinsted, Thule institute, Oulu
  University, Finland
• Svetlana Jevrejeva, Proudman Oceanographic
  Laboratory, UK
• John C. Moore, Arctic Centre, University of
  Lapland, Finland

2
Motivation

• In a previous study we analyzed the relationship
  between the NAO/AO and ice conditions in the
  Baltic Sea some of the detected oscillations
  were associated with ENSO signals (2.2-2.8, 3.5,
  5.2-5.7, 13-14) Jevrejeva and Moore, 2001

3
Influence from AO on ice conditions

• AO warm phase
• Warmer and wetter in northern Europe
• below normal Arctic SLP
• enhanced surface westerlies in the north Atlantic
• AO cool phase
• Colder drier N.Europe
• Relatively high Arctic SLP
• Weakened surface westerlies in the North Atlantic

Kay Dewar, John M. Wallace, David W. J. Thompson
davet_at_atmos.colostate.edu
4
Motivation

• In a previous study we analysed the relationship
  between the NAO/AO and ice conditions in the
  Baltic Sea some of the detected oscillations
  were associated with ENSO signals (2.2-2.8, 3.5,
  5.2-5.7, 13-14) Jevrejeva and Moore, 2001
• Several studies indicate that the impacts of ENSO
  are more readily seen in the north Atlantic
  sector during winter than summer, that there is a
  roughly 3-month lag between tropical signal and
  extra-tropical response, and that signal to noise
  ratio is rather low Trenberth, 1997
  Pozo-Vázquez et al., 2001 Cassou and Terray,
  2001.
• These considerations motivate our approach in
  developing novel wavelet approaches and applying
  them to simultaneously extract both the signals
  in noisy datasets, and the phase angle between
  tropical and North Atlantic/Arctic signals.
• In this study we describe the connections between
  time series analysis and nonlinear dynamics,
  discuss signal- to noise enhancement, we discuss
  signals with relatively low contribution to total
  variance, however, detected signals are
  statistically significant (70 noise)

5
Objective
Barents 2002

• The main objective of our study is to compare
  statistically significant components from ENSO
  represented by SOI (and Niño3) and ice conditions
  in the Barents and Baltic seas.

6
Data sets

• Data sets treated here
• SOI autumn index (1856-2000) (Ropelewski and
  Jones, 1987)
• AO winter index (1851-1997) (Thompson and
  Wallace, 1998)
• BMI Maximum annual ice extent on the Baltic Sea
  (1720-2000) (Seinä et al, 2001)
• BarentsE April ice extent in the (10E-70E)
  Barents Sea for the period 1864-1998 Vinje,
  2001.
• GreenlandSea April ice extent in the Greenland
  Sea (30W-10E) for the period 1864-1998 Vinje,
  2001.
• Note that SOI and Niño3 data sets are delayed by
  3 months relative to arctic indices.
• Other data sets analyzed but not presented in
  detail
• Time series of date of ice break-up in Riga
  (1708-1990) (Jevrejeva, 2001)
• Time series of date of ice break-up in Helsinki
  (1829-1984) (Leppäranta and Seinä, 1986)
• Winter air temperature in Uppsala (1722-1997)
  (Moberg et al, 1999)

7
Data sets
Nino3
SOI
AO
BMI
BarentsE
Greenland Sea
8
Application of MC-SSA(Monte Carlo Singular
Spectrum Analysis (Allen and Smith, 1996))

• to determine non-linear trends and
  quasi-periodic components (signals) in time
  series of SOI /Niño3 and ice conditions time
  series
• to estimate a significance of the signals by
  comparing results to noise models (only signals
  with 95 of significance level are considered)
• to calculate a contribution from a particular
  signal to the total variance
• to identify and compare significant signals
• to show the evolution of those signals over time

9
SSA
non-linear trend
oscillations
Time series of maximum ice extent in the Baltic
Sea (BMI)
noise
10
Oscillations
Oscillations in time series, rows indicate
the rank of the EOFs, bold figures show
oscillations significant at the 95 against AR(1)
red-noise, others are significant at the 95
level against white noise
SKIP?
11
vinij12
vinij12
1
0.5
0

• Normalized 3.5 yrs signal from SOI (red)
  and from time series of maximum ice extent in the
  Barents Sea (black)

-0.5
-1
-1.5
1850
1900
1950
2000
12
jriga
1
0.5
0

• Normalized 5.7 yrs oscillations from time
  series of date of ice break-up at Riga (Baltic
  Sea, red) and Niño3 (blue)

-0.5
-1
-1.5
1700
1750
1800
1850
1900
1950
2000
13
Results from MC-SSA

• Similar signals, with periods of about 2.2- 2.8,
  3.5, 5.7, and 12.8 years, are detected in time
  series of SOI/Niño3 and ice conditions by
  application of MC-SSA.
• Signals 2.2- 2.8, 3.5, 5.2-5.7 , and 12.8 years
  periodicities are isolated and analysed.
• However the variation over time of amplitude and
  relative phases of the quasi-periodic signals
  seen in the times series has not been
  investigated in any detail, though this is
  clearly needed to aid identification of forcing
  mechanisms.

14
Wavelet transforms

• Wavelet transform (WT) - is a tool for
  analyzing localized variations of power within a
  time series. By application of WT we decompose
  the time series into time-frequency space, in
  order to determine both the dominant modes of
  variability and how those modes vary in time.

The wavelet power spectrum. Contours are
normalized variance, thick black line is the 5
significance level using the red noise model,
solid line indicates the cone of influence.

• Torrence and Compo, A practical guide to
  wavelets, Bull. Amer. Meteor. Soc, 79, 61-78,
  1997.

15
Wavelet transform for two time series

• Continuous wavelet (amplitude and phase, Morlet
  (1985))
• Wavelet power spectrum (a measure of the time
  series variance at each scale (period) and at
  each time)
• Crosswavelet power spectrum
• Wavelet Coherence is used to identify frequency
  bands within which two time series are covarying.

16
AO and ice conditions
BMI strong link to AO BarentsE in 12-16 year
band GreenlandSea almost no significant
coherence with AO. (A common event at
1968) Ice out of phase with AO
AO/GreenlandSea
AO/BarentsE
17
Why does AO have stronger link to Baltic than the
Barents?

• The Baltic is a closed system that is reset every
  year. The Barents Sea is subject to rotation of
  ice in the Arctic and has a longer internal
  memory.

AR1 coefs
18
Wavelet coherency and phase between SOI and AO
contours are wavelet squared coherencies, vectors
indicate the phase difference between the SOI/AO,
mean angle is 358º 8º
SSA normalized components of the 13.9 year
oscillation in winter AO (red) and the 13.5 year
oscillation in autumn SOI (blue) results are
significant at the 95 level against a white
noise model

• The AO exhibits a significant power peak in the
  3.5-5.7 year band between 1935-1950, which is
  also associated with a period of higher power at
  3.5-5.7 years in SOI
• Crosswavelet power and coherence indicate large
  covariance between SOI/AO indices at scales of
  3.5-5.7 years. Furthermore, the coherence phase
  is 358º 8º, showing that SOI and AO signals are
  in phase. (note SOI autumn while AO winter)
• High power and coherence in the AO associated
  with signals on the 12-16 year timescales since
  1940 is linked to the SOI

19
SOI and the Baltic
Phase angle is 2º 8º
Phase angle is 200º 6º
The ENSO signatures found in AO are also present
in the BMI. This confirms the MC-SSA results.
20
Baltic severe ice winters and SOI
The most severe ice conditions in the Baltic Sea
observed over the past 150 years were probably in
1939-40, 1941-42, and 1946-47 Seinä et al.,
2001, all are associated with high power in the
AO SOI 3.5-7.8 year band. The winters of
1887-88, 1888-89 are also in the list of most
severe winters Seinä et al., 2001.
21
SOI and BarentsE
The 12-20 year ENSO signature found in the AO is
also present in the BarentsE. Note Confirms
our results from MC-SSA.
22
MC-SSA focus on the 13-14 year periodicity
Note Phase locked and similar trends in
amplitude. Roots in same physical system?
Phasings? The significance of the EOF pairs
involved in the reconstructions has been tested
against red-noise and we find that the AO pair is
significant at the 98 level, SOI at 83, BMI at
91 and BarentsE at 62.
23
Conclusions

• Significant ENSO signatures found in AO, BMI and
  to some degree BarentsE. (2.2- 2.8, 3.5, 5.2-5.7,
  and 13 years) (Wavelets MC-SSA)
• Severe ice conditions in the Baltic Sea linked to
  AO via signals of 2.2-7.8 year periodicity. Which
  in turn are linked to warm events in the tropical
  Pacific Ocean where similar signals are seen 3
  months earlier.
• Theres a common 13-14 year period in AO, SOI,
  BMI and BarentsE. SOI seems to lead AO by 2
  years. What is the physical mechanism?

24
Acknowledgements

• Financial assistance was provided by the Thule
  Institute and the Academy of Finland. Some of
  our software includes code originally written by
  C. Torrence and G. Compo that is available at
  http//paos.colorado.edu/research/wavelets/ and
  by E. Breitenberger of the University of Alaska
  which were adapted from the freeware SSA-MTM
  Toolkit http//www.atmos.ucla.edu/tcd/ssa.

Aslak Grinsted, ag_at_glaciology.net
25
A 3 month ENSO-Arctic link?

• The mechanism of QBO signal propagation is
  described by Baldwin et al. 2001 the QBO
  modulates extra-tropical wave propagation,
  affecting breakdown of the wintertime
  stratospheric polar vortices. The polar vortex in
  the stratosphere affects surface weather patterns
  providing a mechanism for the QBO to have an
  effect on high latitude weather patterns, and
  hence winter ice severity.

26
Selected references Allen, M.R., and L. A.
Smith, Monte Carlo SSA, detecting irregular
oscillations in the presence of coloured noise,
J. Clim., 9, 3383-3404, 1996. Baldwin, M.P., L.J
Gray, T.J Dunkerton, K. Hamilton, P.H. Haynes,
W.J. Randel, J.R. Holton, M.J. Alexander, I.
Hirota, T. Horinouchi, D.B.A Jones, J.S.
Kinnerslay, C. Marquardt, K. Sato, and M.
Tarahashi, The Quasi-Biennial Oscillation,
Reviews of Geophysics, 39, 170-229, 2001. Ghil,
M., M. R. Allen, M. D. Dettinger, K. Ide, D.
Kondrashov, M. E. Mann, A. W. Robertson, A.
Saunders, Y. Tian, F. Varadi, and P. Yiou,
Advanced spectral methods for climatic time
series, Rev. Geophys., 40(1), 1003,
doi10.1029/2000RG000092, 2002. Jevrejeva, S.,
and J.C. Moore, Singular Spectrum Analysis of
Baltic Sea ice conditions and large-scale
atmospheric patterns since 1708, Geophys. Res.
Lett., 28, 4503-07, 2001. Jevrejeva, S.,
Severity of winter seasons in the northern Baltic
Sea during 1529-1990 reconstruction and
analysis, Clim. Res., 17, 55-62,
2001. Jevrejeva, S., Association between the ice
conditions in the Baltic Sea and the North
Atlantic Oscillation, Nordic Hydrol., 33,
319-330, 2002. Jevrejeva, S., J. Moore, A.
Grinsted. Influence of the Arctic Oscillation and
ENSO on Ice Conditions in the Baltic Sea the
Wavelet Approach. J. Geophys. Res., Atm,
10.1029/2003JD003417, 2003 Moritz, R.E., C.M.
Bitz, and E.J. Steig, Dynamics of Recent Climate
Change in the Arctic, Science, 297, 1497-1502,
2002. Seinä, A., H. Grönvall, S. Kalliosaari,
and J. Vainio, Ice seasons 1996-2000 in Finnish
sea areas / Jäätalvet 1996-2000 Suomen
merialueilla, Meri, Report Series of the Finnish
Institute of Marine Research, 43, 2001. Torrence,
C., and G.P. Compo, A practical guide to wavelet
analysis, Bull. Am. Meteorol. Soc., 79, 61-78,
1998. Vinje T., Barents Sea ice edge variation
over the past 400 years, Extended abstract,
Workshop on Sea-Ice Charts of the Arctic,
Seattle, WA, World Meteorological Organization,
WMO/TD No. 949, 4-6, 1999. Vinje, T., Anomalies
and trends of sea ice extent and atmospheric
circulation in the Nordic Seas during the period
1864-1998, J. Clim., 14, 255-267, 2001.