teleconnections of atlantic multidecadal oscillation sergey kravtsov university of...

30
Teleconnections of Atlantic Multidecadal Oscillation Sergey Kravtsov University of Wisconsin-Milwaukee Department of Mathematical Sciences Atmospheric Science Group Collaborators: M. Wyatt , University of Colorado, USA, A. A. Tsonis , K. Swanson , C. Spannagle , University of Wisconsin-Milwaukee, USA Presentation at A. M. Obukhov Institute of Atmospheric Physics, Moscow, Russia November 17, 2011 http://www.uwm.edu/ kravtsov/

Upload: coleen-thompson

Post on 29-Jan-2016

223 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Teleconnections of Atlantic Multidecadal Oscillation Sergey Kravtsov University of Wisconsin-Milwaukee Department of Mathematical Sciences Atmospheric

Teleconnections of Atlantic Multidecadal Oscillation

Sergey Kravtsov

University of Wisconsin-MilwaukeeDepartment of Mathematical Sciences

Atmospheric Science Group

Collaborators:

M. Wyatt, University of Colorado, USA, A. A. Tsonis, K. Swanson, C. Spannagle, University of Wisconsin-Milwaukee, USA

Presentation at A. M. Obukhov Institute of Atmospheric Physics, Moscow, Russia

November 17, 2011

http://www.uwm.edu/kravtsov/

Page 2: Teleconnections of Atlantic Multidecadal Oscillation Sergey Kravtsov University of Wisconsin-Milwaukee Department of Mathematical Sciences Atmospheric

My background and research interests

• 1993 — MIPT, MS: Singular barotropic vortex on a beta-plane (G. M. Reznik, MS advisor)

• 1998 — FSU, PhD: Coupled 2-D THC/sea-ice models (W. K. Dewar, PhD advisor)

• 1998–2005 — UCLA, PostDoc: Atmospheric regimes, wave–mean-flow interaction, coupled ocean–atmosphere modes (M. Ghil, post-doc advisor; A. Robertson, J. C. McWilliams,

P. Berloff, D. Kondrashov)

Page 3: Teleconnections of Atlantic Multidecadal Oscillation Sergey Kravtsov University of Wisconsin-Milwaukee Department of Mathematical Sciences Atmospheric

2005–present — University of Wisconsin-Milwaukee (UWM), Dept. of Math. Sci., Atmospheric Science group:

• Multi-scale climate variability: atmospheric synoptic eddies/LFV (S. Feldstein, S. Lee, N. Schwartz, J. Peters), oceanic mesoscale turbulence/large-scale response (W. Dewar, A. Hogg, P. Berloff, I. Kamenkovich, J. Peters)• Model reduction (D. Kondrashov, M. Ghil, A. Monahan, J. Culina)

• Weather/climate predictability, decadal prediction

• Regional climates and global teleconnections (C. Spannagle, A. Tsonis, K. Swanson, M. Wyatt, P. Roebber, J. Hanrahan)

Page 4: Teleconnections of Atlantic Multidecadal Oscillation Sergey Kravtsov University of Wisconsin-Milwaukee Department of Mathematical Sciences Atmospheric

Topics to be considered:

• Atlantic Multidecadal Oscillation and Northern Hemisphere’s climate variability (with M. Wyatt and A. A. Tsonis)

• Empirical model of decadal ENSO variability

Page 5: Teleconnections of Atlantic Multidecadal Oscillation Sergey Kravtsov University of Wisconsin-Milwaukee Department of Mathematical Sciences Atmospheric

ATLANTIC MULTIDECADAL OSCILLATION AND NORTHERN

HEMISPHERE’S CLIMATE VARIABILITY

M. G. Wyatt, S. Kravtsov, and A. A. Tsonis

(Published in Climate Dynamics, April 2011)

Page 6: Teleconnections of Atlantic Multidecadal Oscillation Sergey Kravtsov University of Wisconsin-Milwaukee Department of Mathematical Sciences Atmospheric

–0.5ºC +0.5ºC

Leading EOF of the

difference between

CMIP-3 multimodel

ensemble mean and observed surface

temperature (2008,

(Kravtsov and Spannagle)

• Dominated by anomalies in North Atlantic region

• Has a multi-decadal timescale

• Has been identified in GCMs as an intrinsic mode

Page 7: Teleconnections of Atlantic Multidecadal Oscillation Sergey Kravtsov University of Wisconsin-Milwaukee Department of Mathematical Sciences Atmospheric

Network of climate indices• NHT — surface air temperature in the NH

• AMO — Atlantic Multi-decadal Oscillation

• AT (AC) — Atmospheric mass Transfer (or Atmospheric Circulation) Index

• NAO — North Atlantic Oscillation

• PDO — Pacific Decadal Oscillation

• NPO — North Pacific Oscillation

• ALPI — Aleutian Low Pressure Index

Page 8: Teleconnections of Atlantic Multidecadal Oscillation Sergey Kravtsov University of Wisconsin-Milwaukee Department of Mathematical Sciences Atmospheric

Preliminary analysis• 13-yr running-mean filtered indices

• lagged correlations found between pairs of climate indices

• Statistical significance of lagged correlations and compatible pairs of indices:

3 yr 3 yr 2 yr

5 yr 5 yr = 3 yr + 2 yr: Compatible indices

Page 9: Teleconnections of Atlantic Multidecadal Oscillation Sergey Kravtsov University of Wisconsin-Milwaukee Department of Mathematical Sciences Atmospheric

M-SSA on our annual climate-index network

• significance estimates based on uncorrelated

red-noise fits to members of index network

• M-SSA — analogous to EOF analysis, but uses, additionally, lagged covariance info

Page 10: Teleconnections of Atlantic Multidecadal Oscillation Sergey Kravtsov University of Wisconsin-Milwaukee Department of Mathematical Sciences Atmospheric

Reconstructed Components:

• Each index is de-

composed into multi-

decadal signal (blue)

and higher-frequency

variability (red)

• Extended 15-index

network

• Relative variations of

the two are to scale

Page 11: Teleconnections of Atlantic Multidecadal Oscillation Sergey Kravtsov University of Wisconsin-Milwaukee Department of Mathematical Sciences Atmospheric

Multidecadal Signal: Stadium Wave

Page 12: Teleconnections of Atlantic Multidecadal Oscillation Sergey Kravtsov University of Wisconsin-Milwaukee Department of Mathematical Sciences Atmospheric

Summary for Stadium Wave• The NH climate indices exhibit a multi-decadal signal inconsistent with random alignment of uncorrelated red-noise time series

•This stadium-wave signal has the following phase relationships (lags in yr, uncertainties estimated using bootstrap re-sampling of index subsets):

• Modeling studies provide clues to the dynamics behind the stadium-wave links

Page 13: Teleconnections of Atlantic Multidecadal Oscillation Sergey Kravtsov University of Wisconsin-Milwaukee Department of Mathematical Sciences Atmospheric

Multidecadal Pacing of Interannual Deviations From the Stadium Wave

• Consider the anomalies with respect to the stadium-wave signal (red lines on an earlier Fig.)

• Fit a multi-dimensional red-noise model that mimics the climatological lag-0, and lag-1 auto- and cross-correlations among the indices

• Compute (almost) the sum of squared cross- correlations for various subsets of indices over sliding window of 5–10 yr: connectivity measure

• Identify index subsets and years with abnormal connectivity values exceeding those expected from the red-noise model

Page 14: Teleconnections of Atlantic Multidecadal Oscillation Sergey Kravtsov University of Wisconsin-Milwaukee Department of Mathematical Sciences Atmospheric

Identification of synchronizing index subsets in 6-index subnets

191719231940

1958

1976

Yellow/orange cells indicate abnormal synchronizations within 6-index subsets

Page 15: Teleconnections of Atlantic Multidecadal Oscillation Sergey Kravtsov University of Wisconsin-Milwaukee Department of Mathematical Sciences Atmospheric

Identification of synchronizing index subsets in 6-index subnets

1917

1940

1976

“Successful” synchronizations were followed by a climate shift (Tsonis, Swanson, Kravtsov 2007)

Page 16: Teleconnections of Atlantic Multidecadal Oscillation Sergey Kravtsov University of Wisconsin-Milwaukee Department of Mathematical Sciences Atmospheric

Climate shifts are characterized by change of dominant climate pattern over the NH (e.g. the

1976 shift) and by different NAO & ENSO regime

1940 1976

Strong ENSO/

NAO

Weak ENSO/

NAO

Strong ENSO/

NAO

Page 17: Teleconnections of Atlantic Multidecadal Oscillation Sergey Kravtsov University of Wisconsin-Milwaukee Department of Mathematical Sciences Atmospheric

Discussion• A multi-decadal climate signal is tentatively generated in the North Atlantic Ocean due to intrinsic variability of the MOC (THC)

• This signal “propagates” across the entire NH as a sequence of delayed teleconnections — stadium wave

• The stadium wave is associated with climate regime shifts which alter the character of interannual climate variability (ENSO and NAO)

• The dynamical processes behind regime shifts may themselves feed back onto and pace the stadium wave

Page 18: Teleconnections of Atlantic Multidecadal Oscillation Sergey Kravtsov University of Wisconsin-Milwaukee Department of Mathematical Sciences Atmospheric

AN EMPIRICAL MODEL OF DECADAL ENSO VARIABILITY

S. Kravtsov

(Submitted to Climate Dynamics)

Page 19: Teleconnections of Atlantic Multidecadal Oscillation Sergey Kravtsov University of Wisconsin-Milwaukee Department of Mathematical Sciences Atmospheric

Conjecture: Modulation of ENSO activity is due to “stadium wave” teleconnections

• Consider seasonal sea-surface temperature (SST) time series on a 5x5º grid (30ºS–60ºN) during 20th century

• Use spatiotemporal filter to isolate multidecadal signal!

Examples: EOFs (Preisendorfer 1988), M-SSA (Ghil

et al. 2002), OPPs (DelSole 2001, 2006), DPs (Schneider and

Held 2001), APT (DelSole and Tippett 2009a,b).

• Despite multidecadal and interannual variability

have different spatial patterns, which vary

according to their respective predominant time scales,

they may still be dynamically linked!

Page 20: Teleconnections of Atlantic Multidecadal Oscillation Sergey Kravtsov University of Wisconsin-Milwaukee Department of Mathematical Sciences Atmospheric

SST discriminants• Patterns that maximize ratio of multidecadal to interannual SST variance (Schneider and Held 2001); SST data is based on Kaplan (1998).

• Time series

correlated

with global Ts

• This and

next pattern

~AMO+PDO

Page 21: Teleconnections of Atlantic Multidecadal Oscillation Sergey Kravtsov University of Wisconsin-Milwaukee Department of Mathematical Sciences Atmospheric

Multidecadal variations in Niño-3• Niño-3 SST is natu-

rally dominated by

interannual variability

(DPs’ contribution is

small)

• Niño-3 variance

exhibits multidecadal

modulation anti-correlated with the AMO index (cf. Federov and

Philander 2000; Dong and Sutton 2005; Dong et al. 2006;

Timmermann et al. 2007)

Page 22: Teleconnections of Atlantic Multidecadal Oscillation Sergey Kravtsov University of Wisconsin-Milwaukee Department of Mathematical Sciences Atmospheric

Niño-3 modulation an artifact?• Due to random sampling (Flügel et al. 2004)• CVs themselves are largely the long-term

modulation of ENSO

Analysis Procedure:

• Generate surrogate SST time series using

multivariate linear inverse modeling (LIM)• Decompose surrogate SSTs into CVs and anomalies, regress Niño-3 STD onto three leading

compute correlation between actual andcompare with observed

CVs,reconstructed Niño-3 STD,correlation

Page 23: Teleconnections of Atlantic Multidecadal Oscillation Sergey Kravtsov University of Wisconsin-Milwaukee Department of Mathematical Sciences Atmospheric

Conclusion: Correlation btw large-scale predictors and ENSO is unlikely to be due to random factors

RESULTS

Page 24: Teleconnections of Atlantic Multidecadal Oscillation Sergey Kravtsov University of Wisconsin-Milwaukee Department of Mathematical Sciences Atmospheric

Let’s model this process statistically• Model Niño-3 index x as a 1-D stochastic process

where f is a polynomial function of x with coefficients

that depend on time t (seasonal cycle) and external

decadal variables y given by leading Canonical Variates

(CV) of SST; dw is a random deviate.

• Study the numerical and algebraic structure of

this model and use it to estimate potential predictability

of decadal ENSO modulations

dx=f(x,y,t)dt+dw

Page 25: Teleconnections of Atlantic Multidecadal Oscillation Sergey Kravtsov University of Wisconsin-Milwaukee Department of Mathematical Sciences Atmospheric

Properties of the empirical ENSO model-I

Page 26: Teleconnections of Atlantic Multidecadal Oscillation Sergey Kravtsov University of Wisconsin-Milwaukee Department of Mathematical Sciences Atmospheric

Properties of the empirical ENSO model-II

Page 27: Teleconnections of Atlantic Multidecadal Oscillation Sergey Kravtsov University of Wisconsin-Milwaukee Department of Mathematical Sciences Atmospheric

Algebraic structure of ENSO model

dx=f(x,y,t)dt+dw; f≡-∂F/∂x

•F – potential function

Page 28: Teleconnections of Atlantic Multidecadal Oscillation Sergey Kravtsov University of Wisconsin-Milwaukee Department of Mathematical Sciences Atmospheric

Cross-validated hindcasts of ENSO STD:

• Jack-knifing with 15-yr segments omitted/predicted

• Linearly extrapolated or fixed external predictors (fixed better!)

• 2 or 3 external predictors (2 better!)

Page 29: Teleconnections of Atlantic Multidecadal Oscillation Sergey Kravtsov University of Wisconsin-Milwaukee Department of Mathematical Sciences Atmospheric

Summary

• These results argue that decadal ENSO modulations are potentially predictable, subject to

our ability to forecast AMO-type climate modes.

• We used statistical SST decomposition into multidecadal and interannual components to define low-frequency predictors (CVs).

• An empirical Niño-3 model trained on the entire 20th-century SST data and forced by CVs captures a

variety of observed ENSO characteristics, including

multidecadal modulation of ENSO intensity.

• The cross-validated hindcasts using linear extrapolation of external predictors are promising

Page 30: Teleconnections of Atlantic Multidecadal Oscillation Sergey Kravtsov University of Wisconsin-Milwaukee Department of Mathematical Sciences Atmospheric

THANKS FOR YOUR ATTENTION