synchronization in intermittent turbulence and...
TRANSCRIPT
sid.inpe.br/mtc-m19@80/2010/04.29.12.38-TDI
SYNCHRONIZATION IN INTERMITTENT
TURBULENCE AND SPATIOTEMPORAL CHAOS IN
THE SOLAR TERRESTRIAL ENVIRONMENT
Rodrigo Andres Miranda Cerda
Doctorate Thesis at Post Graduation Course in Space Geophysics, advised by Drs.
Abraham Chian Long-Chian, and Erico Luiz Rempel, approved in May 01, 2010.
URL of the original document:
<http://urlib.net/8JMKD3MGP7W/37DH6JE >
INPE
Sao Jose dos Campos
2010
PUBLISHED BY:
Instituto Nacional de Pesquisas Espaciais - INPE
Gabinete do Diretor (GB)
Servico de Informacao e Documentacao (SID)
Caixa Postal 515 - CEP 12.245-970
Sao Jose dos Campos - SP - Brasil
Tel.:(012) 3208-6923/6921
Fax: (012) 3208-6919
E-mail: [email protected]
BOARD OF PUBLISHING AND PRESERVATION OF INPE INTEL-
LECTUAL PRODUCTION (RE/DIR-204):
Chairperson:
Dr. Gerald Jean Francis Banon - Coordenacao Observacao da Terra (OBT)
Members:
Dra Inez Staciarini Batista - Coordenacao Ciencias Espaciais e Atmosfericas (CEA)
Dra Maria do Carmo de Andrade Nono - Conselho de Pos-Graduacao
Dra Regina Celia dos Santos Alvala - Centro de Ciencia do Sistema Terrestre (CST)
Marciana Leite Ribeiro - Servico de Informacao e Documentacao (SID)
Dr. Ralf Gielow - Centro de Previsao de Tempo e Estudos Climaticos (CPT)
Dr. Wilson Yamaguti - Coordenacao Engenharia e Tecnologia Espacial (ETE)
Dr. Horacio Hideki Yanasse - Centro de Tecnologias Especiais (CTE)
DIGITAL LIBRARY:
Dr. Gerald Jean Francis Banon - Coordenacao de Observacao da Terra (OBT)
Marciana Leite Ribeiro - Servico de Informacao e Documentacao (SID)
Deicy Farabello - Centro de Previsao de Tempo e Estudos Climaticos (CPT)
DOCUMENT REVIEW:
Marciana Leite Ribeiro - Servico de Informacao e Documentacao (SID)
Yolanda Ribeiro da Silva Souza - Servico de Informacao e Documentacao (SID)
ELECTRONIC EDITING:
Viveca Sant´Ana Lemos - Servico de Informacao e Documentacao (SID)
sid.inpe.br/mtc-m19@80/2010/04.29.12.38-TDI
SYNCHRONIZATION IN INTERMITTENT
TURBULENCE AND SPATIOTEMPORAL CHAOS IN
THE SOLAR TERRESTRIAL ENVIRONMENT
Rodrigo Andres Miranda Cerda
Doctorate Thesis at Post Graduation Course in Space Geophysics, advised by Drs.
Abraham Chian Long-Chian, and Erico Luiz Rempel, approved in May 01, 2010.
URL of the original document:
<http://urlib.net/8JMKD3MGP7W/37DH6JE >
INPE
Sao Jose dos Campos
2010
Cataloging in Publication Data
Miranda Cerda, Rodrigo Andres.M672s Synchronization in intermittent turbulence and spatiotempo-
ral chaos in the solar terrestrial environment / Rodrigo AndresMiranda Cerda. – Sao Jose dos Campos : INPE, 2010.
153 p. ; (sid.inpe.br/mtc-m19@80/2010/04.29.12.38-TDI)
Thesis (Doctorate Thesis in Spatial Geophysics) – NationalInstitute For Space Research, Sao Jose dos Campos, 2010.
Advisers : Drs. Abraham Chian Long-Chian, and Erico LuizRempel.
1. Synchronization. 2. Turbulence. 3. Spatiotemporal chaos.4. Intermittency. 5. Coherents structures. I.Tıtulo.
CDU 523.62-726
Copyright c© 2010 do MCT/INPE. Nenhuma parte desta publicacao pode ser reproduzida, ar-mazenada em um sistema de recuperacao, ou transmitida sob qualquer forma ou por qualquermeio, eletronico, mecanico, fotografico, reprografico, de microfilmagem ou outros, sem a permissaoescrita do INPE, com excecao de qualquer material fornecido especificamente com o proposito deser entrado e executado num sistema computacional, para o uso exclusivo do leitor da obra.
Copyright c© 2010 by MCT/INPE. No part of this publication may be reproduced, stored in aretrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying,recording, microfilming, or otherwise, without written permission from INPE, with the exceptionof any material supplied specifically for the purpose of being entered and executed on a computersystem, for exclusive use of the reader of the work.
ii
To Amélia, my father Eduardo, my mother María Cecilia,my brother Eduardo and my sister Carolina,
and to all the victims of the 2010 Chilean earthquake.
ACKNOWLEDGEMENTS
First of all, I would like to thank my beloved girlfriend Amelia Naomi Onohara, for
her love and patience, for staying with me during sunny days, under the rain and
during storms.
I would like to thank my supervisors, Prof. Abraham Chian-Long Chian, and Prof.
Erico Luiz Rempel, for their patience, advice, incentive and valuable friendship.
I thank Prof. Michio Yamada, Dr. Yoshitaka Saiki and RIMS of Kyoto University
for their kind hospitality.
To Coordenacao de Aperfeicoamento de Pessoal de Nıvel Superior (CAPES) and
Conselho Nacional de Desenvolvimento Cientıfico e Tecnologico (CNPq), for finan-
cial support.
To Dr. Daiki Koga, Dr. Felix Borotto, and Mr. Pablo Munoz for their advice and
friendship.
I would like to thank Dr. Ricardo Monreal and Ms. Cecilia Llop for their incentive
to enter the PhD course.
To Dr. Ezequiel Echer, Prof. Roberto Bruno, Prof. Kristoff Stasiewicz, Dr. Olga
Alexandrova, Prof. Melvin Goldstein, Prof. Bruce T. Tsurutani and Prof. Charles
Meneveau for stimulating discussions.
To the Cluster FGM and CIS instrument teams, the ACE MAG and SWEPAM
instrument teams, the SOHO FGM instrument team, Prof. Fernando Ramos, Dr.
Mauricio Bolzan and Prof. Nalin B. Trivedi for providing the data used in this thesis.
And at last but not least, I would like to thank all my colleagues and friends at
INPE: Aline, Marlos, Valentin, Wanderson, Mauricio, Sergio, Fabio and Yang, and
my friends I met in Japan: Azusa, Inubushi, Ichiyama, Mauricio, Fabricio, Fabian
and the rest of the “latino mafia” at Kyoto University. Thanks for your friendship.
ABSTRACT
In this work we analyze synchronization due to multiscale interactions in obser-vations of intermittent turbulence and numerical simulations of spatiotemporal in-termittency in neutral fluids and space plasmas. This study is carried out in twoparts. First, we apply two distinct nonlinear techniques, kurtosis and phase coher-ence index, to measure the degree of non-Gaussianity and phase synchronizationof intermittent magnetic field turbulence observed in the ambient solar wind, inthe solar photosphere and in the ground, and intermittent atmospheric turbulenceobserved in the Amazon rain forest canopy. Next, we analyze a spatially-extendedmodel of nonlinear waves in fluids and plasmas to identify transient coherent struc-tures responsible for the on-off spatiotemporal intermittency observed in the timeseries of energy. We quantify the degree of amplitude-phase synchronization usingthe power-phase spectral entropy at the onset of spatiotemporal chaos. The observa-tional and theoretical results indicate that the amplitude-phase synchronization maybe the origin of intermittency in fully-developed turbulence in the solar-terrestrialenvironment.
SINCRONIZACAO EM TURBULENCIA INTERMITENTE E CAOSESPACO-TEMPORAL NO AMBIENTE SOLAR-TERRESTRE
RESUMO
Neste trabalho de Tese analisamos a sincronizacao devido a interacoes entre escalasem observacoes de turbulencia intermitente e em simulacoes numericas de inter-mitencia espaco-temporal em fluidos neutros e plasmas espaciais. Este estudo e feitoem duas partes. Primeiro, aplicamos duas tecnicas nao-lineares, curtose e ındice decoerencia de fase, para medir o grau de nao-Gaussianidade e sincronizacao de faseda turbulencia de campo magnetico intermitente observada no vento solar, na foto-sfera solar e no solo, e da turbulencia atmosferica intermitente observada na copada floresta Amazonica. Depois, analisamos um modelo espacialmente estendido deondas nao-lineares em fluidos e plasmas para identificar estruturas coerentes tran-sientes, responsaveis pela intermitencia on-off espaco-temporal observada nas seriestemporais da energia. Quantificamos o grau de sincronizacao de amplitude e faseusando a entropia espectral de potencia e de fase no regime logo depois da transicaopara caos espaco-temporal. Os resultados observacionais e teoricos indicam que asincronizacao de amplitude e fase pode ser a origem da intermitencia na turbulenciacompletamente desenvolvida no ambiente solar-terrestre.
CONTENTS
Pag.
LIST OF FIGURES
LIST OF TABLES
LIST OF ABBREVIATIONS
1 INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2 FUNDAMENTALS OF INTERMITTENT TURBULENCE
AND SPATIOTEMPORAL INTERMITTENCY . . . . . . . . . 29
2.1 Concepts of intermittent turbulence . . . . . . . . . . . . . . . . . . . . . 29
2.1.1 Kolmogorov’s 1941 theory . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.1.2 Taylor hypothesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.1.3 Higher-order structure functions . . . . . . . . . . . . . . . . . . . . . . 32
2.1.4 Phase coherence index . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.2 Concepts of spatiotemporal intermittency . . . . . . . . . . . . . . . . . 35
2.2.1 Numerical detection of chaotic saddles . . . . . . . . . . . . . . . . . . 37
2.2.2 Mathematical representation of wave variables . . . . . . . . . . . . . . 38
2.2.3 Fourier-Lyapunov decomposition . . . . . . . . . . . . . . . . . . . . . 42
2.3 Synchronization of chaotic oscillators . . . . . . . . . . . . . . . . . . . . 45
3 OBSERVATION OF SYNCHRONIZATION IN INTERMIT-
TENT TURBULENCE . . . . . . . . . . . . . . . . . . . . . . . . 51
3.1 Synchronization in magnetic field turbulence . . . . . . . . . . . . . . . . 53
3.1.1 non-ICME event . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
3.1.2 ICME event . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
3.2 Synchronization in atmospheric turbulence . . . . . . . . . . . . . . . . . 79
3.3 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
4 THEORY OF SYNCHRONIZATION IN SPATIOTEMPORAL
INTERMITTENCY . . . . . . . . . . . . . . . . . . . . . . . . . . 91
4.1 Spatiotemporal intermittency and chaotic saddles in the Benjamin-Bona-
Mahony (BBM) equation . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
4.2 Synchronization in the BBM equation . . . . . . . . . . . . . . . . . . . 102
5 CONCLUSION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
REFERENCES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
7 APPENDIX A - SYNCHRONIZATION IN THE SOLAR PHO-
TOSPHERE BEFORE AND AFTER A SOLAR FLARE EVENT135
8 APPENDIX B - KOLMOGOROV 1941 THEORY AND ITS EX-
TENSION TO MAGNETOHYDRODYNAMICS . . . . . . . . . 139
8.1 Neutral fluids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
8.2 Magnetized flows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
9 APPENDIX C - SHANNON ENTROPY . . . . . . . . . . . . . . 145
10 APPENDIX D - WAVE ENERGY IN THE BENJAMIN-BONA-
MAHONY EQUATION . . . . . . . . . . . . . . . . . . . . . . . . 149
11 LIST OF PUBLICATIONS . . . . . . . . . . . . . . . . . . . . . . 153
LIST OF FIGURES
Pag.
2.1 Generation of a phase-randomized surrogate and a phase-correlated sur-
rogate from the original data set . . . . . . . . . . . . . . . . . . . . . . . 34
2.2 Maximum transversal Lyapunov exponent λ⊥ as a function of the cou-
pling parameter ε . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
2.3 Projections of chaotic orbits for ε = 0.1 corresponding to (a) the first
coupled Rossler oscillator, (b) the second coupled Rossler oscillator and
(c) the second coupled Rossler oscillator. (d) The same orbit projected
on the (x1, x2) plane. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
2.4 Projections of chaotic orbits for ε = 0.025 of (a) the first coupled Rossler
oscillator, (b) the second coupled Rossler oscillator and (c) the second
coupled Rossler oscillator. (d) The same orbit projected on the (x1, x2)
plane. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.1 Orbit trace of Cluster and spacecraft position of ACE from 19:40:40 UT
on 1 February 2002 to 03:56:38 UT on 3 February 2002 . . . . . . . . . . 56
3.2 Cluster-1 magnetic field and ion bulk flow velocity during the quasi-
perpendicular shock crossing on Julian day 32, 2002, and the quasi-
parallel shock crossing on Julian day 34, 2002. . . . . . . . . . . . . . . . 57
3.3 GOES-10 X-ray fluxes from 28 January 2002 to 5 February 2002 . . . . . 58
3.4 ACE and Cluster-1 magnetic field and plasma parameters for the selected
time interval. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
3.5 Time series of the modulus of magnetic field of Cluster-1 and ACE, after
removing the trend. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
3.6 Power spectral density (PSD) of |B| for Cluster-1 and ACE, and Com-
pensated PSD for Cluster-1 and ACE. . . . . . . . . . . . . . . . . . . . 62
3.7 Power spectral density of |B| for Cluster-1 and ACE, and their confidence
intervals. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
3.8 Scale dependence of the normalized magnetic field-differences of Cluster
and ACE. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
3.9 The integrand of Equation 3.2, for p = 0 and p = 4, determined from the
magnetic field fluctuations of Cluster-1 and ACE. . . . . . . . . . . . . . 67
3.10 Variations of structure functions with timescale τ calculated from the
magnetic field fluctuations of Cluster-1 and ACE, and structure functions
after applying the Extended Self-Similarity technique. . . . . . . . . . . . 68
3.11 Scaling exponent ζ of the p-th order structure function for Cluster-1 and
ACE magnetic field fluctuations. . . . . . . . . . . . . . . . . . . . . . . 70
3.12 Kurtosis and phase coherence index of |B| measured by Cluster-1 and
ACE. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
3.13 SOHO MDI solar image, kurtosis and the phase coherence index of AR
09802 and a quiet region. . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
3.14 SOHO MDI solar image, kurtosis and the phase coherence index of AR
10720 and a quiet region. . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
3.15 Time series of |B| measured by ACE and Cluster on 1-3 February 2002,
and kurtosis and the phase coherence index of |B|. . . . . . . . . . . . . 76
3.16 Time series of |B| measured by ACE on 21-22 January 2005, and kurtosis
and the phase coherence index of |B|. . . . . . . . . . . . . . . . . . . . . 76
3.17 Time series, kurtosis and the phase coherence index of |B| measured by
ACE, and |B| of the Earth’s geomagnetic field measured by a ground
magnetometer at Ji-Parana on 1-3 February 2002. . . . . . . . . . . . . . 77
3.18 Time series |B| measured by ACE, and time series, kurtosis and phase
coherence index of the modulus of the Earth’s geomagnetic field measured
by a ground magnetometer at Vassouras on 1-3 February 2002. . . . . . . 77
3.19 Time series of temperature and vertical wind velocity above and within
the Amazon canopy. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
3.20 Scale dependence of the normalized temperature-difference and the nor-
malized vertical wind velocity-difference above the Amazon forest canopy. 81
3.21 Scale dependence of the normalized temperature-difference and the nor-
malized vertical wind velocity-difference within the Amazon forest canopy. 82
3.22 PDF of the normalized vertical wind velocity-difference and the normal-
ized temperature-difference above the Amazon forest canopy. . . . . . . . 83
3.23 PDF of the normalized vertical wind velocity-difference and the normal-
ized temperature-difference within the Amazon forest canopy. . . . . . . 84
3.24 Kurtosis and phase coherence index of vertical wind velocities and tem-
peratures above and within the Amazon forest canopy. . . . . . . . . . . 85
3.25 Kurtosis and phase coherence index of vertical wind velocities and tem-
peratures above and within the Amazon forest canopy. . . . . . . . . . . 86
4.1 Spatiotemporal patterns of the regularized long wave equation for ε =
0.199, and ε = 0.201. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
4.2 Time-averaged power spectra in the k wavenumber domain for ε = 0.199
and ε = 0.201. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
4.3 Time-averaged power spectral entropy as a function of the driver ampli-
tude ε. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
4.4 Time series of wave energy E and the maximum peak value of power
spectrum h. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
4.5 Projections of attracting and non-attracting chaotic sets for the regular-
ized long wave equation, for ε = 0.199 and ε = 0.201. . . . . . . . . . . . 101
4.6 Average duration of laminar intervals τ as a function of the departure
from the critical value of the control parameter. . . . . . . . . . . . . . . 102
4.7 Amplitude and phase dynamics of the spatiotemporally chaotic attractor
after the crisis-like transition (ε = 0.20005). . . . . . . . . . . . . . . . . 104
4.8 Three-dimensional projections of attracting and non-attracting chaotic
sets for the regularized long wave equation, for ε = 0.199 and ε = 0.20005.105
4.9 Amplitude and phase dynamics of the temporally chaotic saddle and the
spatiotemporally chaotic saddle after the crisis-like transition. . . . . . . 106
4.10 Time-averaged power spectra and phase-difference spectra of STCA,
STCS and TCS. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
4.11 Spectrum of positive rescaled Lyapunov exponents, power spectral en-
tropy and phase spectral entropy as a function of Lyapunov index j. . . . 109
4.12 Time-average of power-Lyapunov spectra and phase-Lyapunov spectra of
STCA, STCS and TCS. . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
4.13 Comparison of dissipation terms of the Benjamin-Bona-Mahony equa-
tion, and the Shell model of turbulence . . . . . . . . . . . . . . . . . . . 113
7.1 Hinode SOT G-band image taken at 01:00:32 UT on 13 December 2006. 136
7.2 Kurtosis and the phase coherence index as a function of spatial scale r
computed from AR 10930 and a quiet region before the flare. . . . . . . . 137
7.3 Kurtosis and the phase coherence index as a function of spatial scale r
computed from AR 10930 before and after the solar flare. . . . . . . . . . 137
9.1 Decomposition of three possibilities p1 = 1/2, p2 = 1/3 and p3 = 1/6 into
two possibilities with probability 1/2. If the second occurs then there is
another choice with probabilities 2/3 and 1/3. . . . . . . . . . . . . . . . 145
LIST OF TABLES
Pag.
3.1 Numerical examples of flatness for three time scales . . . . . . . . . . . . 66
4.1 Kolmogorov-Sinai entropy and Kaplan-Yorke dimension of the spatiotem-
porally chaotic saddle and temporally chaotic saddle ε = 0.21 . . . . . . 110
LIST OF ABBREVIATIONS
ACE – Advanced Composition ExplorerBBM – Benjamin-Bona-Mahony equationCME – Coronal Mass EjectionESS – Extended Self-SimilarityICME – Interplanetary Coronal Mass EjectionIMF – Interplanetary Magnetic FieldK41 – Kolmogorov’s 1941 TheoryKdV – Korteweg-de Vries equationLBA – Large-scale Biosphere-Atmosphere ExperimentODE – Ordinary Differential EquationORG – Original datasetPCS – Phase Coherent SurrogatePDE – Partial Differential EquationPDF – Probability Density FunctionPRS – Phase Randomized SurrogateRLWE – Regularized Long-Wave EquationSTC – Spatiotemporal ChaosSTCA – Spatiotemporally Chaotic AttractorSTCS – Spatiotemporally Chaotic SaddleTC – Temporal ChaosTCA – Temporally Chaotic AttractorTCS – Temporally Chaotic Saddle
1 INTRODUCTION
Turbulence in neutral fluids can be described as “a spatially complex distribution of
eddies which are advected in a chaotic manner” (DAVIDSON, 2004). There are few
exact results in turbulence theory. Maybe the most famous result is the theory pre-
sented by Kolmogorov (1941). Starting from the Navier-Stokes equations describing
the dynamics of incompressible fluids, and assuming homogeneity and isotropy, the
following equation can be obtained:
⟨(δu)3
⟩= −4
5εr (1.1)
where δu = u(x + r) − u(x), u is a component of the fluid velocity, ε is the mean
energy dissipation rate, r represents spatial scale and 〈〉 represent the ensemble
mean, or the ensemble average. Scale r is assumed to be smaller than the scale of
energy injection (L) into the fluid, and greater than the scale in which molecular
effects become important (η). The scale interval η r L is known as the inertial
subrange. From Equation (1.1) it is possible to obtain other important results. For
instance, the power spectrum has a spectral index equal to −5/3 within the inertial
subrange:
E(k) ∝ k−5/3 (1.2)
The solar wind is a radially expanding plasma flux of solar origin, which forms
a cavity in the interstellar space called the heliosphere. During its expansion, the
solar wind acquires turbulent characteristics which in some aspects are similar to
neutral fluid (hydrodynamic) turbulence. Due to the presence of a magnetic field
convected by the solar wind, low-frequency fluctuations can be described by the
magnetohydrodynamic (MHD) theory.
The solar-terrestrial environment provides a natural laboratory for observing inter-
mittent turbulence in space plasmas (KAMIDE; CHIAN, 2007). Power spectra of veloc-
ity and magnetic field fluctuations have spectral indexes near −5/3 (MATTHAEWS et
al., 1982), similar to turbulence in neutral fluids. Hence, one can use statistical tools
traditionally used for studying hydrodynamic turbulence, for the characterization of
intermittent turbulence in space plasmas.
25
In 1941, Kolmogorov suggested that within the inertial subrange, neutral fluid tur-
bulence has self-similar behavior, i.e., there is an homogeneous distribution of en-
ergy among scales, which implies an absence of coherent structures. Observational
evidence indicates that fluctuations of the fluid velocity in neutral fluids and fluctu-
ations of the magnetic field in the solar wind plasma are not self-similar, due to the
presence of inhomogeneities or coherent structures (FRISCH, 1995; BISKAMP, 2003).
Nonlinear energy cascade (direct and inverse) due to multiscale interactions leads to
localized regions in neutral fluids and space plasmas where phase synchronization
(phase coherence) involving a finite degree of phase coupling among a number of
active modes take place. Large amplitude phase coherent structures seen in these
localized regions dominate the statistics of fluctuations at small scales and have
typical lifetime longer than that of incoherent (random-phase) fluctuations in the
background. Large-amplitude coherent structures are responsible for non-Gaussian
probability density functions (PDFs), displaying sharp peaks and fat tails (leptokur-
tic shape). This departure from Gaussian PDFs becomes more pronounced at smaller
scales.
In analytical modeling and numerical simulations of nonlinear systems based on a
set of deterministic equations, chaos theory allows us to describe some phenomena
related to turbulence, such as coexistence of regular and irregular motion, coexis-
tence of coherence and incoherence, broadband power spectra and intermittency.
The analysis of infinite-dimensional dynamical systems modeled by partial differ-
ential equations provide a bridge between chaos theory and fluid dynamics. Such
systems may exhibit a wealth of regimes, which include temporal chaos, character-
ized by patterns which vary chaotically in time but are regular in space, and spa-
tiotemporal chaos in which the dynamics is chaotic in time and irregular in space.
Theoretical studies of nonlinear waves show that phase synchronization associated
with multiscale interactions is the origin of bursts of coherent structures in fully-
developed spatiotemporal chaos in plasmas and neutral fluids (HE; CHIAN, 2003; HE;
CHIAN, 2005).
This Thesis is organized as follows. In Chapter 2 we review some important concepts
on intermittent turbulence and intermittent spatiotemporal chaos. In Chapter 3 we
apply two nonlinear techniques, kurtosis and the phase coherence index, to measure
the degree of non-Gaussianity and phase synchronization in intermittent magnetic
field turbulence observed in the solar photosphere, the interplanetary solar wind and
26
the Earth’s geomagnetic field, and in intermittent atmospheric turbulence observed
in the Amazon rain forest canopy. In Chapter 4 we use a model of nonlinear waves
in fluids and plasmas to identify the transient coherent structures (chaotic saddles)
which are responsible for the on-off spatiotemporal intermittency observed in the
time series of the energy, and quantify the degree of amplitude-phase synchronization
in the laminar (on-state) and bursty (off-state) regimes. The conclusion is presented
in Chapter 5.
27
2 FUNDAMENTALS OF INTERMITTENT TURBULENCE AND
SPATIOTEMPORAL INTERMITTENCY
In this Chapter we review some important concepts on intermittent turbulence
and intermittent spatiotemporal chaos. In Section 2.1 we give an overview of Kol-
mogorov’s 1941 theory, one of the few exact results on turbulence. Then, we explain
the Taylor hypothesis which allows us to analyze turbulence in the temporal domain.
We finalize Section 2.1 presenting the higher-order structure functions and the phase
coherence index. In Section 2.2, after a brief definition of chaos in ordinary differen-
tial equations and partial differential equations we review two numerical algorithms
for the detection of nonattracting chaotic sets, or chaotic saddles. Next, we revise
the Fourier decomposition, and four indexes which will be used in our numerical
simulations to quantify the dynamics of Fourier amplitudes and phases, namely
the Fourier power spectral entropy, the amplitude disorder parameter, the Fourier
phase spectral entropy, and the phase disorder parameter. Finally, we present the
Fourier-Lyapunov decomposition which allows us to get a complete picture of the
correspondence between the Fourier wavenumbers and the Lyapunov wavevectors
basis.
2.1 Concepts of intermittent turbulence
2.1.1 Kolmogorov’s 1941 theory
The dynamics of incompressible fluids can be described by the Navier-Stokes equa-
tions
∂tu + (u · ∇)u = −∇p+ ν∆u + f , (2.1)
∇ · u = 0. (2.2)
where u = u(x, t) denotes the fluid velocity which depends on position x and time
t. Let us define the second and third-order two-point differences as
29
S2(r) =⟨(δu)2
⟩,
S3(r) =⟨(δu)3
⟩,
where δu(r, t) = u(x + r, t)− u(x, t), and 〈〉 represent the ensemble mean, which is
defined as the mean value over all possible values of its argument and can be thought
as the mean value of a large number of measurements carried out in several similar
experiments (MONIN; YAGLOM, 1971). S2 and S3 are also called the second and third-
order structure functions. Assuming homogeneity (i.e., the statistical quantities are
independent of position in space) and isotropy (i.e., there is no preferred direction of
fluid motion), in the limit ν → 0, Kolmogorov (1941) obtained the following relation
S3(r) = −4
5εr, (2.3)
where r = |r| represents spatial scale, and ε represents the mean energy dissipation
rate. The details of this derivation can be found in Appendix B (Chapter 8). By
assuming that turbulence is self-similar at small scales, i. e., there exists an exponent
α such that
δu(λr) = λαδu(r), (2.4)
substituting (2.4) into (2.3) we obtain
λ3αS3 = −4
5ελr, (2.5)
hence α = 1/3.
Eq. (2.3) can be generalized to structure functions of order p
Sp(r) = 〈(δu)p〉 . (2.6)
30
From the self-similarity assumption we can infer that, if S3 ∝ r for p = 3, then in
general Sp ∝ rαp = rp/3, and since (εr)p/3 has exactly the same dimensions as Sp for
p = 3, the structure function of order p should obey (FRISCH, 1995)
Sp = Cpεp/3rp/3, (2.7)
where Cp is a dimensionless constant. For p = 3, Cp = −4/5.
The second-order structure function is related with the energy spectrum by (DAVID-
SON, 2004)
⟨(δu(r))2⟩ ∼ ∫ ∞
π/r
E(k)dk, (2.8)
where E(k) represents the energy of eddies of size r ∼ π/k. Combining Eq. (2.8)
with Eq. (2.7) with p = 2, and taking the derivative with respect to k one can obtain
E(k) ∼ C ′2ε2/3k−5/3, (2.9)
where C ′2 = −(2C2π2/3)/3.
2.1.2 Taylor hypothesis
In an experimental setup (for instance, a flow in a channel or a pipe), the temporal
variations of the fluid velocity detected by a probe immersed in a fluid can be
interpreted as spatial variations in the frame of reference of the mean flow
r = Uτ, (2.10)
where U represents the modulus of the mean velocity of the flow. In the case of
homogeneous isotropic turbulence, U can be taken as the mean velocity of the largest
eddies of scale L, i.e., U =√〈δu(L)2〉 (BOHR et al., 1998). A fluid element at point
x will be at x + r after a time delay τ = r/U . Taylor (1938) conjectured that the
statistics of two-point differences in space
31
δu(r, t) = u(x + r, t)− u(x, t) (2.11)
are equivalent to the statistics of two-point differences in time
δu(x, τ) = u(x, t+ τ)− u(x, t). (2.12)
Hence, Taylor hypothesis allows us to infer the two-point statistics of u in space
from time measurements of u, at one point x. In the following we drop the spatial
argument of δu, unless explicitly stated.
The statistical properties of two-point differences can be investigated by constructing
histograms. To facilitate comparison between histograms obtained from different
datasets, one can subtract the mean value 〈δu〉 and divide each datapoint by its
standard deviation σ =√∑N
i=1(δu(τ)− 〈δu〉)2/(N − 1)
∆u(τ) =δu(τ)− 〈δu〉
σ. (2.13)
Here 〈〉 denote time average, which for a long series should converge to the ensemble
average (MONIN; YAGLOM, 1971). By using (2.13), the histogram will be normalized
(i.e., its standard deviation will be equal to 1) and centered at zero. Normalized
histograms are called probability density functions (PDFs).
2.1.3 Higher-order structure functions
The theoretical and empirical definition of the structure functions in the temporal
domain (i.e., after assuming Taylor hypothesis) are (DE WIT, 2004)
Sp(τ) =
∫ ∞−∞
P (δu(τ))(δu(τ))pdu, (2.14)
Sp,τ =1
N
N∑i=1
(δui)p, (2.15)
32
where P corresponds to the value of the PDF for δu(τ), δui = ui+τ − ui, and N
corresponds to the number of points in the dataset. Different values of the exponent
p give different information about the shape of the PDF. For example, p = 0 gives
the sum of all probabilities which is equal to 1 by definition, p = 1 gives the mean
value of u which should be equal to zero, p = 2 gives the variance of u which is
equal to 1. p = 3 gives a measure of the degree of asymmetry (or skewness) of the
distribution, and p = 4 quantifies the flatness of the distribution. In general, one has
for p ≥ 3 that odd values of p quantify asymmetry and even values of p quantify the
flatness of the distribution. For p > 4, even values of p are also called hyperflatness.
If the PDF follows a Gaussian distribution, the skewness is equal to zero because
the Gaussian is a symmetric function, and the flatness is constant and equal to 3.
Intermittency can be quantified by calculating the normalized fourth-order structure
function K (kurtosis). One can empirically estimate K by
K(τ) =1
N
N∑i=1
(δui − 〈δui〉
σ
)4
− 3, (2.16)
which is equivalent to the flatness minus 3 (FRISCH, 1995). Since the flatness is
given by the velocity fluctuations raised to the fourth power, then both flatness
and kurtosis can be regarded as the kinetic energy squared (i.e., S4,τ = 〈(δui)4〉 =
〈[(δui)2]2〉).
2.1.4 Phase coherence index
An alternative method of quantifying intermittency and non-Gaussianity is to apply
the phase coherence technique using surrogate data by defining a phase coherence
index based on the null hypothesis (HADA et al., 2003; KOGA; HADA, 2003; KOGA et
al., 2007; KOGA et al., 2008; NARIYUKI; HADA, 2006; CHIAN et al., 2008; NARIYUKI et
al., 2008; SAHRAOUI, 2008; TELLONI et al., 2009)
Cφ(τ) =SPRS(τ)− SORG(τ)
SPRS(τ)− SPCS(τ), (2.17)
where
33
FIGURE 2.1 - Generation of a phase-randomized surrogate (PRS, yellow) and a phase-correlated surrogate (PCS, green) from the original (ORG, red) data set.The power spectrum is kept the same, but the phases of Fourier modes ofPRS are all set as random numbers, and the phases of Fourier modes ofPCS are all set to zero.
SOURCE: Adapted from Koga and Hada (2003)
Sj(τ) =n∑i=1
|Bi+τ −Bi|, (2.18)
with j = ORG, PRS, PCS. This index measures the degree of phase synchronization
in an original data set (ORG) by comparing it with two surrogate data sets created
from the original data set: a phase-randomized surrogate (PRS) in which the phases
of the Fourier modes are all set as random numbers, and a phase-correlated surrogate
(PCS) in which the phases of the Fourier modes are all set to the same value. The
power spectrum of three data sets ORG, PRS and PCS are kept the same, but their
phase spectra are different (see Figure 2.1). An average of over 100 realizations of
the phase shuffling is performed to generate the phase-randomized surrogate data
set SPRS(τ). Cφ(τ)=0 indicates that the phases of the scales τ of the original data
are completely random (i.e., null phase synchronization), whereas Cφ(τ)=1 indicates
34
that the phases are fully correlated (i.e., total phase synchronization).
2.2 Concepts of spatiotemporal intermittency
Consider the following set of coupled ordinary differential equations (ODEs)
du(t)
dt= F(u(t)) (2.19)
where F = F1, F2, ..., FN is a coupled nonlinear set of functions of u =
u1, u2, ..., uN. The time variable is a continuous and implicit variable of Eq. (2.19),
and its time evolution (i.e., the continuous set of solutions of Eq. (2.19)) is called
trajectory, orbit or flux. The space spanned by the variables u = u1, u2, ..., uNis called the phase space of system (2.19). The Poincare plane, or Poincare surface
of section, is a “plane” defined in phase space (for example, u1 = cte) which inter-
sects the system’s trajectory in a particular direction (for example, from u1 < cte to
u1 > cte). The definition of a Poincare plane allows us to simplify the analysis of a
N -dimensional continuous set of solutions into a (N − 1)-dimensional discrete set of
solutions. This discrete set of points lying on the Poincare plane are called Poincare
points.
The Lyapunov exponents are the mean separation (or contraction) rate along or-
thogonal directions between two trajectories whose initial conditions are separated
by a very small distance. By sorting the Lyapunov exponents in decreasing order we
can associate the first Lyapunov exponent with the direction of maximum “stretch-
ing”, or minimum contraction. A more detailed definition of Lyapunov exponents
is given in subsection 2.2.3. Here, it is enough to note that if the first Lyapunov
exponent is positive, then the distance between two orbits increase with time, and
if it is negative, then two orbits will tend to be closer with time.
A chaotic trajectory is a trajectory of Eq. (2.19) which satisfies the following condi-
tions (ALLIGOOD et al., 1996)
a) The sequence of Poincare points is not asymptotically periodic (i.e., there
is no periodicity as t→∞).
b) At least one Lyapunov exponent is positive.
35
Now we will define chaotic set and chaotic attractor. A point u(t) in phase space
is an ω-limit point of an initial condition u0 if for any neighborhood V of u, the
trajectory starting from u0 enters in V repeteadly when t → ∞. The set of all ω-
limit points of u0 is called the ω-limit set ω(u0). If u belongs to a chaotic trajectory,
and also belongs to its own ω-limit set ω(u), then this set is called a chaotic set. If
ω(u) is an attracting set, then ω(u) is a chaotic attractor.
Chaos theory can describe some phenomena related to turbulence, such as co-
existence of regular and irregular motion, co-existence of coherence and incoherence,
broadband power spectra, and intermittency. However, the lack of spatial informa-
tion in systems described by a small number of coupled ODEs makes it hard to draw
conclusions on their usefulness for the interpretation of the dynamics of real fluids.
Now consider the following partial differential equation (PDE)
Du(x, t)
Dt= F (u(x, t)), (2.20)
where D/Dt indicates the total time derivative, and F is a nonlinear function of
u. The analysis of infinite-dimensional dynamical systems modeled by PDEs can
provide a bridge between chaos theory and fluid dynamics. Such systems may exhibit
a wealth of regimes, which include temporal chaos (TC) and spatiotemporal chaos
(STC). In PDEs, we refer to temporal chaos whenever the patterns generated vary
chaotically in time, but spatial coherence is preserved. In spatiotemporal chaos,
the dynamics is chaotic in time and irregular in space. Sometimes, the TC and STC
behaviors are referred to as spatiotemporal chaos and fully developed spatiotemporal
chaos, respectively (TEL; LAI, 2008). In relation to turbulence, a comparatively small
number of degrees of freedom is active in spatiotemporal chaos, so the system lacks
a fully developed turbulent cascade (OUELLETTE; GOLLUB, 2008).
In dissipative spatiotemporal systems chaotic dynamics can appear in the form of
asymptotic or transient chaos. Asymptotic chaos refers to the dynamics on chaotic
attractors, while transient chaos is caused by nonattracting chaotic sets known
as chaotic saddles in phase space (KANTZ; GRASSBERGER, 1985; HSU et al., 1988;
BRAUN; FEUDEL, 1996; REMPEL et al., 2004). The coupling of distinct chaotic saddles
embedded in a chaotic attractor results in intermittent switching between transient
states. The coupling between a temporally chaotic saddle (TCS) and a spatiotem-
36
porally chaotic saddle (STCS) has been shown to be responsible for the TC-STC
intermittency in a spatiotemporally chaotic attractor (STCA) (REMPEL; CHIAN,
2007; REMPEL et al., 2007), right after the onset of STC. Chaotic saddles can also be
used to predict the dynamics of the STCA after the onset of spatiotemporal chaos
(REMPEL; CHIAN, 2007).
2.2.1 Numerical detection of chaotic saddles
Chaotic saddles are nonattracting chaotic sets, hence they cannot be studied by
simply integrating equations forward in time. Here we review two numerical schemes
which were used to detect chaotic saddles. Both schemes rely on the definition of a
restraining region R in phase space containing the chaotic saddle, implying that no
attractors are included in R.
2.2.1.1 The sprinkler algorithm
The sprinkler algorithm (HSU et al., 1988) works by first defining a restraining region
R in the Poincare surface of section in which the chaotic saddle lies, then covering
R with a grid of initial conditions, and finally iterating each initial condition until
some time Tc larger than the average escape time from the restraining region. The
escape time T of an initial condition u0 is defined as the minimum time for which
the n-th crossing between the orbit and the Poincare section un is not in R. The final
points which remain in the restraining region approximate the unstable manifold,
their initial conditions approximate the stable manifold, and the points obtained
at T = ξTc approximate the chaotic saddle. For most systems ξ = 0.5 (HSU et
al., 1988; REMPEL et al., 2004). The sprinkler method can be easily implemented
in low- and high-dimensional systems, and is useful for computing the stable and
unstable manifolds of the chaotic saddle (besides the chaotic saddle itself), but some
parameters such as ξ have to be obtained via trial-and-error. The computation of
statistical quantities such as Lyapunov exponents of chaotic saddles can be done
using this algorithm (OTT, 1993), but it is not very precise since the sprinkler method
does not obtain arbitrarily long continuous trajectories.
2.2.1.2 The stagger-and-step algorithm
The stagger-and-step method (SWEET et al., 2001) finds a trajectory which always
stays in the vicinity of a chaotic saddle. It can be implemented as follows. First, select
any initial condition u0 within the restraining region R and a minimum required
37
escape time T ∗. Denote by δ = ||r|| the magnitude of the perturbation vector r.
Randomly perturb the initial condition using an arbitrary δ > 0 until a trajectory
having lifetime T (u0 +r) ≥ T ∗ is found. Set u′0 = u0 +r. Evolve u′0 until the lifetime
of the current point T (u) < T ∗. Then, perturb the current point in phase space
using a small perturbation δ, until T (u + r) ≥ T ∗, set u′ = u + r, and continue
iterating using u′ as initial condition.
In the stagger-and-step method, the choice of the distribution of the random pertur-
bation r is an important aspect. Sweet et al. (2001) suggest using the “exponential
stagger distribution”, which is generated as follows. Let a be such that 10−a = δ, and
let amax be the maximum value of a allowed by the numerical precision available.
Generate a uniformly distributed random number s between a and amax. Choose
a random unit vector x from a uniform distribution on a set of unit vectors. The
random perturbation vector r is obtained by
r = 10−sx. (2.21)
The stagger-and-step method generates a pseudotrajectory (i.e., a trajectory with
numerical precision of the order of δ) which stays in the vicinity of the chaotic
saddle for an arbitrarily long time. Hence, it can be used for the computation of
statistical quantities which require enough datapoints to ensure convergence, such
as the spectrum of Lyapunov exponents.
2.2.2 Mathematical representation of wave variables
Here we present a brief review of the Fourier decomposition and the Fourier-
Lyapunov decomposition, as well as the different indexes which quantify amplitude
and phase dynamics.
2.2.2.1 Fourier representation
For a given wave variable u(x, t) in real space, we expand it in a Fourier series as
u(x, t) =N∑
k=−N
uk(t)eikx, uk(t) ∈ C, (2.22)
38
where uk(t) represents the complex Fourier coefficients
uk(t) =1
N
N∑k=−N
u(x, t)e−ikx, (2.23)
where k = 2πn/L, n = −N, ..., N and L represents the spatial length of the system. If
u(x, t) is a real function, then u−k(t) = u∗k(t), where ∗ denotes the complex conjugate,
hence only wavenumbers k = 1, ..., N have to be considered. From each coefficient
one can extract its amplitude and phase
|uk(t)| =√uk(t) · u∗k(t), (2.24)
φk(t) = arctan
(Im(uk(t))
Re(uk(t))
). (2.25)
2.2.2.2 Fourier power spectral entropy
The power spectral entropy index is the Shannon entropy (SHANNON, 1949) ap-
plied to the amplitude information of Fourier modes, and is defined as (POWELL;
PERCIVAL, 1979; XI; GUNTON, 1995; CAKMUR et al., 1997)
SAk (t) = −N∑k=1
p(uk(t)) ln[p(uk(t))], (2.26)
where uk(t) represents the complex Fourier coefficient with wavenumber k, and
p(uk(t)) is the relative weight of mode k:
p(uk(t)) =|uk(t)|2∑Nk=1 |uk(t)|2
, (2.27)
and the convention ln[p(uk(t))] = 0 for p(uk(t)) = 0 is used. The derivation of the
Shannon entropy is given in Appendix C (Chapter 9).
The power spectral entropy is a measure of the energy spread among Fourier modes.
SAk will be maximum if the energy is uniformly distributed among modes, and min-
39
imum if all energy is concentrated at a certain wavenumber k.
2.2.2.3 Amplitude disorder parameter
The amplitude disorder parameter (also known as the averaged wave number) is
defined as follows (THYAGARAJA, 1979; MARTIN; YUEN, 1980; HE; CHIAN, 2003)
DAk (t) =
√∑Nk=1 k
2|uk(t)|2∑Nk=1 |uk(t)|2
. (2.28)
This quantity represents the square root of the ratio between the enstrophy k2|uk|2
and the energy |uk|2 (OHKITANI; YAMADA, 1989). It is a measure of the average
number of active modes.
2.2.2.4 Fourier phase spectral entropy
The phase spectral entropy index derived from the Shannon entropy is given by
(POLYGIANNAKIS; MOUSSAS, 1995)
Sp = −N∑k=1
P (φk(t)) ln[P (φk(t))], (2.29)
where P (φk(t)) denotes the probability distribution function (PDF) of the Fourier
phase φk(t), which is determined by constructing a normalized histogram of φk(t).
For P (φk(t)) = 0, ln[P (φk(t))] = 0. This method of detecting phase synchronization
can be improved using the phase difference (TASS et al., 1998; CHIANG; COLES, 2000;
LAI et al., 2006)
δφk(t) = φk+1(t)− φk(t), (2.30)
where δφk is restricted to the [−π, π] interval, due to the cyclic nature of the phase.
Substituting (2.30) into Eq. (2.29) we obtain
Sφk = −N∑k=1
P (δφk(t)) ln[P (δφk(t))]. (2.31)
In the presence of phase synchronization, the PDF of Fourier phase differences
40
P (δφk) will tend to concentrate on a narrow range in δφk and Sφk will tend to zero.
In the absence of synchronization, all phase differences have the same probability
of occurrence, hence P (δφk) will tend to an uniform distribution and Sφk will be
maximum.
2.2.2.5 Phase disorder parameter
The order parameter was originally formulated by Kuramoto (1984) to quantify the
degree of phase synchronization among identical oscillators. It is defined as follows
R(t) =
∣∣∣∣∣ 1
N
N∑k=1
exp iφk(t)
∣∣∣∣∣ , (2.32)
where φk represents the Fourier phases of each oscillator k.
To characterize synchronization in nonidentical oscillators we propose a modification
of the order parameter by using phase differences
Rφ(t) =
∣∣∣∣∣ 1
N
N∑k=1
exp iδφk(t)
∣∣∣∣∣ . (2.33)
The order parameter defined in Eq. (2.33) quantifies the degree of phase synchro-
nization of a set of oscillators which do not need to be identical. In order to facilitate
visual comparison with the power and phase spectral entropies, we define the phase
disorder parameter as follows
Dφk (t) = 1−Rφ(t) = 1−
∣∣∣∣∣ 1
N
N∑k=1
exp iδφk(t)
∣∣∣∣∣ , (2.34)
where Dφ = 0 represents perfect synchronization among oscillators, and Dφ = 1
indicates that the oscillators are completely desynchronized.
41
2.2.3 Fourier-Lyapunov decomposition
From the Fourier decomposition of a wave variable
u(x, t) =N∑
k=−N
uk(t)ekx, uk(t) ∈ C, (2.35)
substituting into Eq. (2.20) we can write a set of ODEs representing the dynamics
of the complex amplitudes uk(t) as
dukdt
= fk(uk). (2.36)
By decomposing each complex amplitudes into real and imaginary parts, uk = uRk +
iuIk, uRk , uIk ∈ R, i =√−1, one can rewrite Equation (2.36) as
du
dt= F(u). (2.37)
Note that the phase space of system (2.37) is a real 2N space uRk , uIk, k = 1, ..., N .
Let us denote by u0 an initial condition of system (2.37), and φt(u0, t0) its solution,
that is (PARKER; CHUA, 1989)
φt(x0, t0) = F(φt(x0, t0), t), φt0(x0, t0) = x0. (2.38)
Taking the derivative of Eq. (2.38) with respect to x0 we obtain
Dx0φt(x0, t0) = DxF(φt(x0, t0), t)Dx0φt(x0, t0), Dx0φt0(x0, t0) = I, (2.39)
were I denotes the identity matrix. Let us define the flux Jacobian Φt(x0, t0) =
Dx0φt(x0, t0). Then Eq. (2.39) becomes
Φt(x0, t0) = DxF(φt(x0, t0), t)Φt(x0, t0), Φt0(x0, t0) = I (2.40)
42
Equation (2.40) is known as the variational equation (PARKER; CHUA, 1989). A small
perturbation δu0 of u0 will evolve as
δu = Φt(x0, t0)δu0 (2.41)
The asymptotic behavior of perturbation δu is given by the Lyapunov spectrum,
which is the set of Lyapunov characteristic exponents λj defined by (SHIMADA;
NAGASHIMA, 1979; YAMADA; SAIKI, 2007)
λ1 + λ2 + ...+ λj = limt→∞
1
t− t0ln
(||δu1(t) ∧ δu2(t) ∧ ... ∧ δuj(t)||||δu1(0) ∧ δu2(0) ∧ ... ∧ δuj(0)||
), (2.42)
where ∧ represents the exterior (wedge) product, double bars denote the norm, and
j = 1, ..., 2N . The Lyapunov exponents defined by equation (2.42) represent the
expanding (or contracting) rate of volume of the j-dimensional parallelepiped in the
tangent space along the orbit having initial conditions δu1(0), δu2(0), ..., δuj(0).
The Kolmogorov-Sinai entropy H can be obtained from the Lyapunov spectrum
H =
q∑j=1
λj, λq > 0, λq+1 ≤ 0, (2.43)
which can be interpreted as a number measuring the time rate of creation of infor-
mation as a chaotic orbit evolves (OTT, 1993). Another useful quantity which can
be obtained from the Lyapunov spectrum is the Kaplan-Yorke dimension
D = p+
p∑j=1
λj
|λp+1|, p = maxm|
m∑j=1
λj ≥ 0. (2.44)
2.2.3.1 Fourier-Lyapunov amplitude and phase dynamics
Following Yamada and Ohkitani (1998), the complex Fourier-Lyapunov vector is a
vector with components
δujk = (δuRk + iδuIk)j, k = 1, ..., N ; j = 1, ..., 2N. (2.45)
43
From each Fourier-Lyapunov vector we can extract information of its amplitude and
phase
∣∣δujk(t)∣∣ =
√[(δuRk (t) + iδuIk(t)) · (δuRk (t) + iδuIk(t))
∗]j,
φjk(t) =
[arctan
(δuIk(t)
δuRk (t)
)]j.
The time-averaged Fourier-Lyapunov power spectrum is given by
⟨∣∣δujk∣∣2⟩ =⟨∣∣(δuRk + iδuIk)
j∣∣2⟩ .
The time-averaged Fourier-Lyapunov phase spectrum is defined as
⟨δφjk⟩
=⟨
(φk+1 − φk)j⟩. (2.46)
Using the Fourier-Lyapunov representation we define the power spectral entropy as
SAj (t) = −N∑k=1
p(δujk(t)) ln[p(δujk(t))], (2.47)
where p(δujk(t)) is the relative weight of Fourier mode k of Lyapunov wavevector j
p(δujk(t)) =|δujk(t)|2∑Nk=1 |δu
jk(t)|2
. (2.48)
We define the phase spectral entropy using the Fourier-Lyapunov representation as
Sφj (t) = −N∑k=1
P (δφjk(t)) ln[P (δφjk(t))], (2.49)
where P (δφjk(t)) denotes the probability distribution function (PDF) of the Fourier-
44
Lyapunov phase difference δφjk(t), which can be determined by constructing a nor-
malized histogram of δφjk(t). For P (δφjk(t)) = 0, ln[P (δφjk(t))] = 0.
2.3 Synchronization of chaotic oscillators
In this section we review some important concepts of synchronization between cou-
pled, chaotic oscillators. Consider the following system of two coupled, nonlinear
identical oscillators:
dx
dt= f(x) + ε · (y − x) (2.50)
dy
dt= f(y) + ε · (x− y), (2.51)
where x ∈ <m, y ∈ <m, f represent a vector field, and ε represents the coupling
parameter. The synchronization manifold is a subspace in which the oscillators are
completely synchronized, i.e., x = y (FUJISAKA; YAMADA, 1983). Let us denote the
synchronization manifold as M . The transverse stability of M can be determined by
introducing the following transform of variables (LAI et al., 2003):
(u,v) =
[1
2(x + y),
1
2(x− y)
], (2.52)
Inserting (2.52) into (2.50) and (2.51) gives:
du
dt+dv
dt= f(u + v)− 2ε · v (2.53)
du
dt− dv
dt= f(u− v) + 2ε · v. (2.54)
Adding eqs. (2.53) and (2.54) gives:
2du
dt= f(u + v) + f(u− v) (2.55)
45
Near the synchronization manifold one has |v| ∼ 0, and the synchronization state is
given by v = 0. Eq. (2.55) then reads:
2du
dt= f(u) + f(u)
=⇒ du
dt= f(u). (2.56)
Now substracting Eq. (2.54) from (2.53):
2dv
dt= f(u + v)− f(u− v)− 4ε · v. (2.57)
Expanding terms f(u + v) and f(u− v) into a Taylor series around u, one has:
f(u + v)|u = f(u) +∂f(u)
∂u(u + v − u) + ... (2.58)
f(u− v)|u = f(u) +∂f(u)
∂u(u− v − u) + ... (2.59)
Inserting (2.58) and (2.59) into (2.57):
2dv
dt= f(u) +
∂f(u)
∂u(u + v − u)− f(u)− ∂f(u)
∂u(u− v − u)− 4ε · v + ... (2.60)
Keeping first-order terms, one has:
2dv
dt≈ 2
∂f(u)
∂uv − 4ε · v
=⇒ dv
dt≈ ∂f(u)
∂uv − 2ε · v
=
[∂f(u)
∂u− 2ε
]v. (2.61)
46
0 0.05 0.1 0.15 0.2ε
-0.4
-0.2
0
0.2
λ ⊥
FIGURE 2.2 - Maximum transversal Lyapunov exponent λ⊥ as a function of the couplingparameter ε
.
From Eq. (2.56) and Eq. (2.61) it is possible to obtain the conditional (or “trans-
verse”) Lyapunov exponents of system (2.50)-(2.51). If the maximum conditional
Lyapunov exponent is negative (positive), then the synchronization manifold is
transversally stable (unstable). Fujisaka and Yamada (1983) introduced a numer-
ical algorithm to obtain the conditional Lyapunov exponents, and applied it to a
system of two coupled Lorenz equations. The bifurcation in which the conditional
Lyapunov exponents changes from negative to positive value is referred to as a
blowout bifurcation (MANSCHER et al., 1998). When the synchronization manifold
becomes unstable, the system can display on-off intermittency (PLATT et al., 1993;
MANSCHER et al., 1998) which is an aperiodic switching between laminar behav-
ior and chaotic bursts, due to orbits entering and leaving every sufficiently small
neighborhood of the synchronization manifold (PLATT et al., 1993).
As an example, consider the following system of three coupled Rossler oscillators
(LAI et al., 2003)
dxi/dt = −ωiyi − zi + ε(xi+1 + xi−1 − 2xi), (2.62)
dyi/dt = ωixi + ayi, (2.63)
dzi/dt = b+ zi(xi − c), (2.64)
47
where i = 1, 2, 3, ωi is the mean frequency of the ith oscillator, ε is the coupling
parameter and a, b and c are parameters of each individual Rossler oscillator. We
choose ωi = ω = 1, a = 0.165, b = 0.2 and c = 1. The transverse (conditional)
Lyapunov exponents of system (2.62)-(2.64) are obtained by solving Eq. (2.61) (i.e.,
the variational equation) which reads
dv
dt=
−3ε −ω −1
ω a 0
z 0 x− c
v. (2.65)
Figure 2.2 shows the maximum conditional Lyapunov exponent λ⊥ of Eqs. (2.62)-
(2.64) as a function of the coupling parameter ε. The value of λ⊥ decreases with
increasing ε, and at εc ∼ 0.065, the sign of λ⊥ changes from positive to negative,
and the synchronization manifold becomes transversely stable.
Figure 2.3(a)-(c) shows the (xi, yi) projection of the chaotic orbits of each Rossler
oscillator (i = 1, 2, 3) for ε = 0.1, corresponding to λ⊥ < 0. The trajectories are
chaotic, but confined to the synchronization manifold as shown in Fig. 2.3(d). For
ε = 0.025, we observe in Fig. 2.4(a)-(c) that trajectories are still chaotic but not
restrained to the synchronization manifold which is no longer stable, since λ⊥ > 0.
48
-20 -10 0 10 20x
1
-20
-10
0
10
20y 1
Rossler 1
-20 -10 0 10 20x
2
-20
-10
0
10
20
y 2
Rossler 2
-20 -10 0 10 20X
3
-20
-10
0
10
20
y 3
Rossler 3
-20 -10 0 10 20x
1
-20
-10
0
10
20
x 2
(a) (b)
(c) (d)
FIGURE 2.3 - Projections of chaotic orbits for ε = 0.1 corresponding to (a) the firstcoupled Rossler oscillator, (b) the second coupled Rossler oscillator and (c)the second coupled Rossler oscillator. (d) The same orbit projected on the(x1, x2) plane.
-20 -10 0 10 20x
1
-20
-10
0
10
20
y 1
Rossler 1
-20 -10 0 10 20x
2
-20
-10
0
10
20
y 2
Rossler 2
-20 -10 0 10 20X
3
-20
-10
0
10
20
y 3
Rossler 3
-20 -10 0 10 20x
1
-20
-10
0
10
20
x 2
(a)
(c)
(b)
(d)
FIGURE 2.4 - Projections of chaotic orbits for ε = 0.025 of (a) the first coupled Rossleroscillator, (b) the second coupled Rossler oscillator and (c) the second cou-pled Rossler oscillator. (d) The same orbit projected on the (x1, x2) plane.
49
3 OBSERVATION OF SYNCHRONIZATION IN INTERMITTENT
TURBULENCE
In this Chapter, the techniques described in Section 2.1 are applied to two exam-
ples of intermittent turbulence observed in the solar-terrestrial environment: the
intermittent magnetic field turbulence observed in the solar wind using data from
satellites, in the solar photosphere using solar magnetograms and in the ground us-
ing magnetometers, and the intermittent atmospheric turbulence observed in the
Amazon rain forest canopy.
The solar wind provides a natural laboratory for observation of intermittent mag-
netic field turbulence (BRUNO; CARBONE, 2005; KAMIDE; CHIAN, 2007). Nonlinear
energy cascade (direct and inverse) due to multi-scale interactions leads to localized
regions of space plasmas where phase synchronization (phase coherence) involving a
finite degree of phase coupling among a number of active modes takes place. Large-
amplitude phase coherent structures seen in these localized regions dominate the
statistics of fluctuations at small scales and have typical lifetime longer than that of
incoherent (random-phase) fluctuations in the background.
A recent theoretical study of nonlinear waves shows that phase synchronization
associated with multi-scale interactions is the origin of bursts of coherent structures
in intermittent turbulence in plasmas and fluids (HE; CHIAN, 2003; HE; CHIAN, 2005).
Observational evidence in support of these findings in space plasma turbulence was
obtained by Hada et al. (2003), Koga and Hada (2003), Koga et al. (2007) and Koga
et al. (2008) using the Geotail solar wind data upstream and downstream of Earth’s
bow shock, by Sahraoui (2008) using the Cluster data in the magnetosheath close
to the Earth’s magnetopause, and by Telloni et al. (2009) using the SOHO data of
solar corona; and in atmospheric turbulence by Chian et al. (2008) using the Amazon
forest data.
Neutral fluid turbulence can be studied experimentally via measurements of the
Earth’s atmospheric turbulence such as the wind velocity or temperature fluctua-
tions. In the latter case, the Amazon rain forest plays a key role in the regional and
global climate dynamics. One important problem for understanding the vegetation-
atmosphere interactions in Amazonia is the turbulent exchange of scalar and mo-
mentum in the atmospheric boundary layer - above and within the forest canopy.
51
Atmospheric turbulence in the Amazon forest has been extensively investigated. For
example, Fitzjarrald et al. (1990) performed detailed observations of turbulence just
above and below the crown of an Amazon forest during the wet season. This analysis
shows that the forest canopy removes high-frequency turbulent fluctuations while
passing lower frequencies. A study of CO2 concentration was made by Sternberg et
al. (1997) in two different forests in the Amazon basin during the dry season, one site
characterized by a closed canopy structure in which turbulent mixing is minimized
and another site characterized by an open canopy structure in which the turbulent
mixing is maximized. This analysis shows that the respiratory CO2 recycling in the
closed canopy forest with lower wind speeds is occurring to a greater extent than
the open canopy forest with higher wind speeds. The vertical dispersion of trace
gases using a Lagrangian approach was analyzed by Simon et al. (2005) based on in-
canopy turbulence data collected at Jaru and Cuieiras Reserves. This study indicates
that for day-time conditions when there is an efficient turbulent mixing in the upper
canopy and profile gradients are small, the radon-222 source/sink distributions show
a high sensitivity to small measurement errors and the CO2 and H2O fluxes show
a reasonable agreement with the eddy covariance measurements made above the
forest canopy, which is not the case for night-time conditions when the CO2 profile
gradients in the upper canopy are large due to reduced turbulent mixing.
The remaining of this Chapter is divided as follows. Section 3.1 is devoted to the char-
acterization of intermittency and phase synchronization in intermittent magnetic
field turbulence. In particular, subsection 3.1.1 aims to seek further observational
evidence of phase synchronization in space plasmas based on the magnetic field data
of Cluster and ACE (Advanced Composition Explorer) spacecraft. We compare the
phase synchronization detected by Cluster in the magnetic field turbulence in the
shocked solar wind upstream of Earth’s bow shock with the phase synchronization
detected by ACE in the magnetic field turbulence in the unshocked ambient solar
wind at the L1 Lagrangian point. In subsection 3.1.2 we study the scale dependence
of kurtosis and phase coherence in intermittent magnetic field turbulence measured
at three different locations of the solar-terrestrial environment: (1) in the solar pho-
tosphere, (2) in the solar wind, and (3) on the ground. We investigate two scenarios:
a non-ICME event in February 2002 and an ICME event in January 2005. Section
3.2 aims to apply the kurtosis (fourth-order structure function) and phase coherence
techniques to determine the intermittent nature of day-time atmospheric turbulence
above and within the Amazon forest canopy. We show that both techniques are ca-
52
pable of characterizing the dissimilarity of scalar and velocity in above-canopy and
in-canopy atmospheric turbulence.
3.1 Synchronization in magnetic field turbulence
3.1.1 non-ICME event
The physical conditions upstream of Earth’s bow shock along the path of Cluster are
expected to differ from the unshocked ambient solar wind in the vicinity of ACE. The
magnetic connection between the interplanetary magnetic field (IMF) and the bow
shock may occur sporadically in the upstream solar wind, as evidenced by a strong
emission at the local electron plasma frequency (KELLOGG; HORBURY, 2005). In
contrast the ambient solar wind at L1, being far away from the Earth’s bow shock,
is not affected by the shock. This Section carries out a comparative study of the
degree of phase synchronization across a wide range of scales in the interplanetary
magnetic field fluctuations in shocked (Cluster) and unshocked (ACE) regions of
solar wind.
Cluster has observed intermittent interplanetary turbulence upstream of Earth’s
bow shock. The first study of solar wind intermittency using Cluster data was re-
ported by Pallocchia et al. (2002). They showed that velocity fluctuations detected
by Cluster-3 are slightly more intermittent than Cluster-1 on 22 February 2001. Bale
et al. (2005b) used the Cluster-4 data of 19 February 2002 to show that both electric-
field and magnetic-field fluctuations of turbulence in the upstream solar wind display
the k−5/3 spectral behavior of classical Kolmogorov fluid turbulence over an inertial
subrange and a spectral break at kρi ∼ 0.45 (where ρi is the ion Larmor radius). In
the dissipative subrange above the spectral break point, the magnetic spectrum be-
comes steeper while the electric spectrum gets enhanced. They suggest that Alfven
waves in the inertial subrange eventually disperse as kinetic Alfven waves above the
spectral break, becoming more electrostatic at short wavelengths where wave energy
is dissipated through wave-particle interaction processes such as Landau or tran-
sit time damping. Narita et al. (2006) determined directly the wavenumber power
spectra of intermittent magnetic field turbulence in the foreshock of a quasi-parallel
bow shock using four-point Cluster spacecraft measurements; they conjectured that
nonlinear interactions of Alfven waves can lead to phase coherence in the foreshock
turbulence observed by Cluster. Alexandrova et al. (2007) used the Cluster-1 data
of 5 April 2001 to demonstrate that in the inertial subrange below the ion cy-
53
clotron frequency, the turbulent spectrum of unshocked solar wind magnetic field
follows Kolmogorov’s law. However, after the spectral break the turbulence cannot
be characterized by a “dissipative” range. Instead, the kurtosis (fourth-order struc-
ture function) increases with frequency, similar to the intermittent behavior of the
low-frequency inertial subrange, indicating that nonlinear wave interactions are op-
erating to yield a new high-frequency inertial subrange. Alexandrova et al. (2008)
showed that the magnetic field fluctuations within the high-frequency inertial sub-
range identified by Alexandrova et al. (2007) is much more compressive than the
low-frequency inertial subrange dominated by incompressive Alfven waves. This in-
crease of compressibility is due to a partial dissipation (and destruction of phase
coherence) of left-hand Alfvenic fluctuations by the ion cyclotron damping in the
neighborhood of the spectral break point around the ion cyclotron frequency, leading
to a new right-hand “magnetosonic” small-scale cascade characterized by an increase
of intermittency as well as spectrum steepening.
ACE has monitored solar wind in an orbit about the L1 point. Burlaga and Vinas
(2004) showed that the fluctuations of solar wind speed observed by ACE are re-
lated to intermittent turbulence and shocks at the smallest scales (1 hour) and can
be described by a Tsallis probability distribution function derived from nonextensive
statistical mechanics. Smith et al. (2006) demonstrated that while the inertial sub-
range of solar wind magnetic field turbulence measured by ACE at lower frequencies
displays a tightly constrained range of spectral indexes, the dissipation range ex-
hibits a broad range of power-law indexes. Chapman and Hnat (2007) showed that
solar wind turbulence detected by ACE is dominated by Alfvenic fluctuations with
power spectral exponents that evolve toward the Kolmogorov value of - 5/3, and
can be decomposed into two coexistent components perpendicular and parallel to
the local average magnetic field. Hamilton et al. (2008) found that on average the
wave vectors of solar wind magnetic field turbulence measured by ACE are more
field-aligned in the dissipation subrange than in the inertial subrange, and cyclotron
damping plays an important but not exclusive role in the formation of the dissipa-
tion subrange; moreover, the orientation of the wave vectors for the smallest scales
within the inertial subrange are not organized by wind speed and that on average the
data shows the same distribution of energy between perpendicular and field-aligned
wave vectors.
Recently, a phase coherence technique for characterizing phase synchronization in
54
nonlinear wave-wave coupling and turbulence based on surrogate data has been de-
veloped for space plasmas (HADA et al., 2003; KOGA; HADA, 2003; KOGA et al., 2008;
SAHRAOUI, 2008). The link between phase coherence, non-Gaussianity and intermit-
tent turbulence was established by Koga et al. (2007), based on the Geotail magnetic
field data upstream and downstream of Earth’s bow shock. In this subsection, we
investigate phase synchronization due to nonlinear multiscale interactions and non-
Gaussian statistics using the magnetic field data collected by Cluster upstream of
Earth’s bow shock and by ACE in the ambient solar wind at L1. By applying the
phase coherence index technique to quantify the degree of phase synchronization,
we show that its variation with time scales is similar to kurtosis indicating a signifi-
cant departure from Gaussianity over a wide range of time scales, which is enhanced
at small scales, in agreement with the leptokurtic shape of small-scale probability
density function (PDF) of intermittent magnetic field fluctuations in both regions
of space plasmas.
3.1.1.1 Cluster and ACE data of 1 to 3 February 2002
Figure 3.1 depicts the orbit trace of Cluster and spacecraft position of ACE, in the
GSE Cartesian coordinate system, from 19:40:40 on 1 February 2002 to 03:56:38
on 3 February 2002 during which Cluster-1 traverses the upstream region of the
Earth’s bow shock. For this time interval, ACE appears practically stationary in
the scales of Figure 3.1 and the Cluster tetrahedron scale (i.e., the mean distance
between spacecrafts) was small (∼ 100-300 km). For spacecraft separations of 300
km and mean solar wind bulk velocity of 374 km/s (obtained for the selected time
interval) and assuming the Taylor’s hypothesis, the time scale above which all 4
Cluster spacecraft observe the same eddies is ∼ 0.8 s. In this study, our analysis
will cover time scales above 1 s (Figure 3.12), hence the differences of measurements
between the satellites are indistinguishable. The selected time interval, defined by
the onset of the solar wind supersonic/subsonic transitions, begins when Cluster-1
crosses the shock front of a quasi-perpendicular bow shock by entering into the solar
wind at the time indicated by a dashed line in the upper panel of Figure 3.2, and
ends when Cluster-1 departs from the solar wind by entering into the transition
(foreshock) region of a quasi-parallel bow shock at the time indicated by a dashed
line in the lower panel of Figure 3.2. In contrast to a quasi-perpendicular shock
(BALE et al., 2005a) characterized by sharp transitions of the modulus of the ion
bulk flow velocity |Vi| and magnetic field |B|, a quasi-parallel shock (BURGESS et
55
160240
0
080
40
80
−40
−80
GSE
GS
E
ACE
ER
RE
X ( )
Y
(
)
Cluster
FIGURE 3.1 - Orbit trace of Cluster and spacecraft position of ACE, in the GSE coor-dinate system, from 19:40:40 UT on 1 February 2002 to 03:56:38 UT on 3February 2002. The starting position of Cluster is shown as a full circle.
SOURCE: Chian and Miranda (2009)
al., 2005) is characterized by a transition region with repeated shock crossings, as
seen in Figure 3.2. This quasi-parallel shock event has been analyzed by a number of
papers (EASTWOOD et al., 2003; STASIEWICZ et al., 2003; BEHLKE et al., 2004; LUCEK
et al., 2004). When the Cluster spacecraft navigate in the upstream solar wind they
stay always very close to the bow shock, as a result magnetic connections to the bow
shock occur frequently (KELLOGG; HORBURY, 2005). Although we have selected an
interval outside of the foreshock region of a quasi-parallel bow shock the magnetic
connection happens from time to time, for example, between 00:50 and 01:00 UT,
and between 01:20 and 01:36 UT on 3 February 2002. Hence, the plasma conditions
of solar wind seen by Cluster-1 are different from that seen by ACE at L1 since the
solar wind turbulence measured by Cluster-1 is a combination of the ambient solar
wind plus fluctuations coming from the bow shock. Note that during the selected
56
0
100
200
300
400
500
0
100
200
300
400
500
0
10
20
30
40
0
10
20
30
40
19:25:40
03:41:38
Quasi−parallel shock crossing
Quasi−perpendicular shock crossing
19:40:40
03:56:38
Time (UT)
Day 32
Day 34
19:55:40
04:11:38
|B|
|B|
|V |
|V |i
i
FIGURE 3.2 - Cluster-1 magnetic field |B| (red, nT) and ion bulk flow velocity |Vi| (black,km/s) during the quasi-perpendicular shock crossing (upper panel) on Ju-lian day 32, 2002, and the quasi-parallel shock crossing (lower panel) onJulian day 34, 2002. The vertical dashed lines indicate the beginning andthe end of the selected time interval of Figure 3.4, respectively.
SOURCE: Chian and Miranda (2009)
interval no M- or X-class solar flares were detected, as evidenced by the GOES-10 X-
ray data shown in Figure 3.3, and strong interplanetary disturbances such as ICMEs
were not seen.
In this study, we perform a nonlinear analysis of the modulus of magnetic field
|B| = (B2x +B2
y +B2z )
1/2. We are interested in analyzing the relation between phase
synchronization and intermittency of solar wind magnetic field turbulence which
does not require a detailed analysis of its field components. As a matter of fact, in a
similar study Bruno et al. (2003) showed that the modulus and the components of the
solar wind magnetic field give the same qualitative behaviors of intermittency. The
Cluster and ACE magnetic fields are detected by the FGM instruments (BALOGH et
al., 2001; SMITH et al., 1998) at a resolution of 22 Hz and 1 Hz, respectively, providing
a set of 2,604,208 and 116,159 data points, respectively, for the interval chosen. For
the sake of completeness, Figure 3.4 presents an overview of other in situ plasma
57
28 30 32 34 36Julian days of year 2002
10-9
10-8
10-7
10-6
10-5
10-4
10-3
Inte
nsity
(W
/m^2
)
X-Class event
M-Class event
1-8 A0.5-4 A
FIGURE 3.3 - GOES-10 X-ray fluxes from 28 January 2002 to 5 February 2002 (Julian day36). Red dashed lines indicates thresholds for X-class (10−4) and M-class(10−5− 10−4) events. None of the M-class X-ray events detected happenedduring the selected time interval.
parameters for this interval. The three components of the vector magnetic field Bx,
By and Bz are given in the GSE coordinates. ΦB and ΘB denote the angle of the
solar wind magnetic field relative to the Sun-Earth x-axis in the ecliptic plane, and
the angle out of the ecliptic, respectively, in the polar GSE coordinates (EASTWOOD
et al., 2003). These angles can be obtained from the following relations
ΦB = tan−1
(By
Bx
),
ΘB = tan−1
(Bz√
B2x +B2
y
).
Figure 3.4 also shows the modulus of the ion bulk flow velocity |Vi|, the ion number
density ni and the ion temperature Ti (where the component perpendicular to the
magnetic field for Cluster is plotted). It follows from Figure 3.4 that during this
time interval Cluster and ACE are immersed in a slow solar wind. The ion plasma
58
05
1015
20
-20-10
010
0
180
360
-90
0
90
300
360
420
0
20
40
104
106
0
10
20
Time (UT)
11:48:39Day 33
19:40:40 11:48:39Day 33
03:56:3803:56:38Day 34 Day 32 Day 34
19:40:40Day 32
|B|
nT
β
ACE
ΦB
ΘB
Cluster
ii
i|V
| iB
xB
yB
z
FIGURE 3.4 - ACE and Cluster-1 magnetic field and plasma parameters for the selectedtime interval. From top to bottom: modulus of magnetic field |B| (nT); thethree components of the vector magnetic field (nT) Bx, By, and Bz in theGSE coordinates; angle ΦB (degrees) of the magnetic field relative to thex axis in the ecliptic plane; angle ΘB (degrees) of the magnetic field outof the ecliptic; ion bulk flow velocity |Vi| (km/s), ion number density ni(cm−3), ion temperature Ti (Kelvin), and ion plasma beta βi.
SOURCE: Chian and Miranda (2009)
59
βi (the ratio between plasma kinetic pressure and magnetic pressure) is calculated
by the following expression
βi =2µ0kBniTi
B20
,
where µ0 = 4π × 10−7 [Henry/m] is the permeability of vacuum, kB = 1.38× 10−23
[Joule/K] is the Boltzmann constant, ni is the ion number density, Ti is the ion
temperature, and B0 is defined here as the mean value of |B|. Plasma measure-
ments from ACE are provided by the Solar Wind Electron Proton Alpha Monitor
(MCCOMAS et al., 1998), while Cluster plasma measurements are given by the Ion
Spectrometry experiment CIS (REME et al., 2001). Note that the CIS instrument of
Cluster-1 is switched from the telemetry mode 14 (Compression Magnetosphere-4)
to the telemetry mode 5 (Compression Solar Wind-4) at 21:55:11 UT on Julian Day
32, and then to the telemetry mode 10 (Magnetosphere-3) at 01:15:04 UT on Ju-
lian Day 34 (REME et al., 2001), which account for the discontinuities seen in the
beginning and at the end of the ion number density and ion temperature profiles
indicated by arrows in Figure 3.4.
Although the interplanetary magnetic field behaves sometimes as a stationary pro-
cess (BRUNO; CARBONE, 2005), in the time interval studied here there is a trend in
the time series of |B| of Cluster and ACE, as seen in Figure 3.4. In order to guarantee
the stationarity of data we remove a trend from |B| by subtracting a cubic fitting
(MACEK et al., 2005) computed from the time series of |B|. Figure 3.5 shows the
resulting stationary time series of |B| for Cluster and ACE, which display sporadic
bursts of large-amplitude spikes typical of intermittency.
The upper panels of Figure 3.6 show the power spectral density (PSD) of |B| for
Cluster and ACE, corresponding to the time series of |B| of Figure 3.5; they depict
a typical power spectrum density of solar wind turbulence with a spectral break
separating the inertial subrange from the dissipative subrange, each with its own
power law (LEAMON et al., 2000; BRUNO et al., 2005; ALEXANDROVA et al., 2008).
The power spectral density was computed using the Welch method (WELCH, 1967),
which consists of dividing the time series into M subintervals, multiplying each
subinterval term by term by a window function (in our case we used the Hanning
window (PASCHMANN; DALY, 2000)) and then computing the power spectrum of each
60
FIGURE 3.5 - Time series of the modulus of magnetic field |B| (nT) of Cluster-1 andACE, after removing the trend by computing a cubic fitting of the originaldata.
SOURCE: Chian and Miranda (2009)
subinterval using the fast Fourier transform. The average of the M power spectra
gives the PSD. This method reduces the error of the spectrum estimate, resulting
in a narrower PSD. The spin frequency fspin of both spacecraft are indicated in
Figure 3.6. The Nyquist frequency fNyq, defined as half of the sampling frequency
fs, fNyq = 0.5fs, which marks the maximum frequency for which the PSD gives
reliable values (PASCHMANN; DALY, 2000), is equal to 11 Hz for Cluster and 0.5 Hz
for ACE.
The frequency range in which each PSD follows a −5/3 Kolmogorov scaling (i.e.,
the inertial subrange) can be determined by constructing the compensated PSD,
multiplying the original PSD by f+5/3 (BISKAMP et al., 1999). The inertial subrange
should appear as a frequency range in which the compensated PSD is almost hori-
zontal (i.e., zero slope). The compensated PSD of Cluster and ACE are shown in the
lower panels of Figure 3.6. To facilitate visualization, each compensated spectrum
is smoothed by dividing it into overlapping subintervals shifted by one datapoint,
each subinterval contains 10 datapoints, and then calculating the mean value within
61
10-6
10-4
10-2
100
102
104
10-3
10-2
10-1
100
10110
-4
10-3
10-2
10-1
10-3
10-2
10-1
100
101
PSD
(nT
/H
z)2
fspin
fNyq
fspin
f (Hz)(Hz)f
PSD
5/3
f
Cluster ACE
−1.6
−2.1
−1.5
−2.0
FIGURE 3.6 - Upper panels: Power spectral density (PSD) of |B| for Cluster-1 and ACE.The spin frequency of each spacecraft is indicated as fspin. The Nyquistfrequency for ACE is indicated as fNyq. Straight lines indicate the inertialand “dissipative” subranges of each spacecraft. The spectral break thatmarks the transition from the inertial to dissipative subrange occurs nearthe local ion cyclotron frequency fci. For both Cluster and ACE fci ∼ 0.12Hz in the solar wind frame, which is Doppler-shifted to a higher frequencyin the spacecraft frame. Lower panels: Compensated PSD for Cluster andACE. To facilitate visualization, a smoothed compensated PSD is shownin black. Vertical dashed lines indicate the beginning and the end of theinertial subranges used to compute the spectral indices of Cluster and ACEin the upper panels.
SOURCE: Chian and Miranda (2009)
62
10-3
10-2
10-1
100
10110
-6
10-4
10-2
100
102
104
10-3
10-2
10-1
100
101
PS
D(n
T /H
z)2
(Hz)f f (Hz)
Cluster ACE
FIGURE 3.7 - Power spectral density of |B| for Cluster-1 and ACE. Black curves indicatethe upper and lower bounds of confidence intervals (error bars).
each subinterval. Each mean value is plotted at the center of the subinterval. The
smoothed compensated spectra of Cluster and ACE are shown in the lower panels of
Figure 3.6 (black curves), and the vertical dashed lines indicate the beginning and
the end of the resulting inertial subrange. For both Cluster and ACE, the beginning
of the inertial subrange is defined as the first value of the smoothed compensated
PSD. For Cluster, the compensated spectrum clearly shows a “knee” (i.e., a local
maximum) to the right of fspin, hence the end of the inertial subrange is defined as
the frequency just before the “knee”. For ACE, the end of the inertial subrange is
defined as the frequency where a change of slope occurs. For Cluster, the inertial
subrange spectral index in the frequency range 0.0026-0.3 Hz is qinert = −1.6± 0.01
and the dissipative subrange spectral index in the frequency range 0.6-3.74 Hz is
qdissip = −2.1± 0.01. For ACE, the inertial subrange spectral index in the frequency
range 0.0013-0.1 Hz is qinert = −1.5 ± 0.01 and the dissipative subrange spectral
index in the frequency range 0.18-0.33 Hz is qdissip = −2.0± 0.04.
The fluctuations of the modulus of magnetic field |B| can be regarded as compressible
(or parallel) fluctuations (SAMSONOV et al., 2007; ALEXANDROVA et al., 2008). The
total power spectral density PSDtotal is closer to the −5/3 Kolmogorov scaling than
the PSD of |B| (PSD‖), where PSDtotal = PSD(Bx)+PSD(By)+PSD(Bz) = PSD⊥+
PSD‖, PSD⊥ denotes the PSD of transverse Alfvenic fluctuations (SAMSONOV et al.,
2007). In the solar wind at 1 AU, magnetic field fluctuations are mostly Alfvenic and
nearly incompressible (ALEXANDROVA, 2008) which implies that, within the inertial
63
subrange, transverse fluctuations contain more power than compressible fluctuations.
For the sake of completeness, we calculated the standard deviation resulting from
averaging a set of power spectra, after applying the Welch method. The standard
deviation of each frequency, also called the confidence intervals, are represented as
black curves in Figure 3.7, superposed by the PSDs of Cluster and ACE.
The spectral break that marks the transition from the inertial to dissipative subrange
occurs near the local ion cyclotron frequency fci = eB/mi in an appropriate frame.
For both Cluster and ACE fci ∼ 0.12 Hz if we use B0 in the solar wind frame.
Since data are taken in the spacecraft frame, fci has to be Doppler-shifted to higher
frequency by a quantity of the order of Vsw/VA, where Vsw is the solar wind speed
and VA is the Alfven speed, in the interpretation of the power spectra in Figure 3.6.
3.1.1.2 Intermittency, non-Gaussianity and phase synchronization
Figure 3.8 shows the scale dependence of the normalized magnetic field-differences
of Cluster and ACE
∆B =δB − 〈δB〉
σB, (3.1)
for three different time scales (τ = 10 s, 100 s and 1000 s), where δB = |B(t+ τ)| −|B(t)| denotes two-point differences of the modulus of magnetic field |B| for a given
time scale (lag) τ , the brackets denote the mean value of δB, and σB denotes the
standard deviation of δB. It is evident from Figure 3.8 that, for both Cluster and
ACE, the magnetic field fluctuations become more intermittent as the scales become
smaller. In terms of spatial scales, the three time scales in Figure 3.8 correspond to
3,740 km, 37,400 km and 374,000 km, respectively, using the mean solar wind (ion
bulk flow) velocity of 374 km/s and assuming the Taylor’s hypothesis. Note that the
spin frequency is filtered from the time series of Cluster and ACE, respectively, by
applying an orthogonal wavelet decomposition to the data using a Daubechies 10
mother wavelet (DAUBECHIES, 1994), and removing the scales corresponding to the
spacecraft spintone of 0.25 Hz for Cluster and 0.083 Hz for ACE.
The intermittent characteristics of interplanetary turbulence can be elucidated by
the probability density function (PDF) of magnetic field fluctuations. The p-th order
of the structure function is formally defined as (DE WIT, 2004)
64
B∆
B∆
Time (UT)
ACE
Cluster
22:59:00Day 33
sτ = 10
sτ = 100
sτ = 10
τ = 1000s
τ = 1000s
τ = 100 s
23:09:30
22:59:00Day 33
23:09:30
23:20:00
23:20:00
FIGURE 3.8 - Scale dependence of the normalized magnetic field-differences of Cluster(red) and ACE (blue) for three different time scales (τ = 10 s, 100 s and1000 s).
SOURCE: Chian and Miranda (2009)
Sp(τ) =
∫ ∞−∞
P (∆B(τ))(∆B(τ))pd(∆B(τ)), (3.2)
where P (∆B) denotes the probability density function (PDF) of magnetic field dif-
ferences ∆B. The first four orders of the structure function are statistical quantities
that characterize PDFs (PAPOULIS, 1965; DAVIDSON, 2004), namely, p = 0 gives the
sum of all probabilities (equal to 1 by definition), p = 1 gives the mean value of
∆B (equal to zero according to Eq. 3.1), p = 2 gives the variance of ∆B (equal to
1 from Eq. 3.1, whose square root is the standard deviation), p = 3 measures the
degree of asymmetry (skewness) of the distribution, and p = 4 quantifies the flatness
of the distribution. Figure 3.9 plots the integrand of Eq. (3.2) for p = 0 (PDF) and
4 (flatness), determined from the magnetic field fluctuations of Cluster and ACE
65
TABLE 3.1 - Numerical examples of flatness for three time scales
τ = 10 s 100 s 1000 s
Cluster 35.86 14.41 6.80
ACE 47.39 16.02 7.67
(Figure 3.5), for 3 different time scales (τ = 10 s, 100 s and 1000 s), superposed
by a Gaussian PDF (grey line). It shows that the PDFs of ∆B for both shocked
and unshocked solar wind are closer to a Gaussian distribution at large time scales
but deviate from a Gaussian distribution as τ decreases. At small scales the shape
of PDF (p = 0) becomes leptokurtic, exhibiting fat tails and sharp peaks. For p
= 4, the flatness of the distribution of ∆B of both Cluster and ACE increases at
small scales, indicating an excess of rare, large-amplitude fluctuations. The areas
spanned by the curves shown in Figure 3.9 for p = 4 approximate the values of the
fourth-order (flatness) structure function (DE WIT, 2004). Table 3.1 lists numerical
examples of flatness which shows that, for all three scales, the level of flatness of
magnetic field fluctuations of ACE is higher than Cluster.
The departure from self-similarity in the magnetic field fluctuations can be quantified
by comparing the scaling exponents of higher-order structure functions within the
inertial subrange against the Kolmogorov universality theory (FRISCH, 1995). The
characterization of departures from Kolmogorov’s 1941 theory (hereafter K41) is
of great interest since the K41 theory is a result based on the assumption that
the turbulence is homogeneous and isotropic (i.e., self-similar). The upper panels of
Figure 3.10 show the structure functions obtained from the following formula (DE
WIT; KRASNOSELSKIKH, 1996),
Sp(τ) = 〈|Bi+τ −Bi|p〉 ∼ τα(p), (3.3)
for p = 1 - 6. The scale is logarithmic for both axes. The grey areas denote the inertial
subranges determined from the power spectral density of Figure 3.6. The scaling
66
10-6
10-3
100
103
Cluster
ACE
10-6
10-3
100
103
-30 -20 -10 0 10 20 3010
-6
10-3
100
103
10-6
10-3
100
103
10-6
10-3
100
103
-30 -20 -10 0 10 20 3010
-6
10-3
100
103
P(∆
B)(∆
B)p
P(∆
B)(∆
B)p
P(∆
B)(∆
B)p
∆B∆B
τ = 100 τ = 100s s
p = 0
τ = 10 s τ = 10 s
τ = 1000 sτ = 1000 s
p = 4
FIGURE 3.9 - The integrand of Equation 3.2, for p = 0 and p = 4, determined fromthe magnetic field fluctuations of Cluster-1 (red) and ACE (blue), for threedifferent time scales (τ = 10 s, 100 s and 1000 s), superposed by a GaussianPDF (grey line). The areas spanned by the curves of Cluster and ACE forp = 4 approximate the value of the flatness.
SOURCE: Chian and Miranda (2009)
67
FIGURE 3.10 - Upper panels: variations of structure functions with timescale τ calculatedfrom the magnetic field fluctuations of Cluster-1 and ACE (upper panels)for p = 1 (black), 2 (purple), 3 (light green), 4 (yellow), 5 (dark green)and 6 (light blue), the grey area indicates the inertial subrange. Lowerpanels: structure functions after applying the Extended Self-Similaritytechnique, the bar indicates the previous inertial subrange, and the greyarea indicates the extended scaling range. For the visualization purpose,the stretched structure functions have been normalized to Sp(T ), whereT = 0.044 s for Cluster and T = 1 s for ACE.
SOURCE: Chian and Miranda (2009)
68
exponent for each order of the structure function can be obtained by estimating the
slope of a linear-fit of the curves within the inertial subrange. In order to improve
the calculation of the scaling exponent, we apply the Extended Self-Similarity (ESS)
technique (BENZI et al., 1993), which consists of plotting each order of the structure
function Sp as a function of S3. This technique allows us to extend the scaling range
where Sp(τ) ∼ [S3(τ)]ζ(p). The scaling exponents ζ(p) ∼ α(p)/α(3) can be found
from the extended range. The lower panels of Figure 3.10 illustrate the application
of this technique. The horizontal line represents the inertial subrange before the
“stretching” process, and the grey area indicates the extended scaling range. Figure
3.11 shows the scaling exponent ζ(p) as a function of p, for Cluster and ACE. The
dashed line denotes the K41 scaling, ζ(p) = p/3. It is evident, from Figure 3.11, that
the scaling exponent measured by both spacecraft display significant departure from
self-similarity, which implies that the magnetic field fluctuations in both regions of
space plasmas are intermittent. For the time interval considered in this work, the
unshocked solar wind magnetic field at L1 is more intermittent than the shocked
solar wind upstream of Earth’s bow shock, in agreement with Table 1.
Intermittency can be quantified by calculating the empirical estimate of the normal-
ized fourth-order moment K (kurtosis) (DE WIT, 2004),
K(τ) =1
n
n∑i=1
(δBi − 〈δBi〉
σB
)4
− 3, (3.4)
which is equivalent to flatness minus 3 (FRISCH, 1995; DAVIDSON, 2004). For a
Gaussian noise K = 0 for all scales; whereas for an intermittent (non-Gaussian)
signal K(τ) > 0 and K increases as scale decreases within the inertial subrange.
Figure 3.12 shows the computed variation of K with the time scale τ for magnetic
field fluctuations of Cluster and ACE. We have chosen the lower bound of 1 s for
Figure 3.12 since the measurements at scales smaller than 1 s are contaminated by
the noise level of FGM instruments. The test to find the smallest time scale is done
by over-plotting the instrument noise level of 10−4 nT2/Hz (for both Cluster and
ACE) on the power spectral density of Figure 3.6.
An alternative method of quantifying intermittency and non-Gaussianity is to apply
the phase coherence technique using surrogate data by defining a phase coherence
index (HADA et al., 2003; KOGA; HADA, 2003; KOGA et al., 2007; KOGA et al., 2008;
69
1 2 3 4 5 6
0.5
1
1.5
2
K41ClusterACE
(p)
ζ
p
FIGURE 3.11 - Scaling exponent ζ of the p-th order structure function obtained by ESSfitting for Cluster-1 and ACE magnetic field fluctuations. The dashed linecorresponds to K41 (self-similar) Kolmogorov scaling.
SOURCE: Chian and Miranda (2009)
NARIYUKI; HADA, 2006; CHIAN et al., 2008; NARIYUKI et al., 2008; SAHRAOUI, 2008;
TELLONI et al., 2009)
Cφ(τ) =SPRS(τ)− SORG(τ)
SPRS(τ)− SPCS(τ), (3.5)
where
Sj(τ) =n∑i=1
|Bi+τ −Bi|, (3.6)
with j = ORG, PRS, PCS. This index measures the degree of phase synchronization
in an original data set (ORG) by comparing it with two surrogate data sets created
from the original data set: a phase-randomized surrogate (PRS) in which the phases
of the Fourier modes are made completely random, and a phase-correlated surrogate
70
0
50
100Cluster
ACE
| |
| |
100
101
102
103
104
0
0.5
1
τ (sec)
Ph
ase
Co
her
ence
In
dex
Ku
rto
sis
cb
a
FIGURE 3.12 - Kurtosis and phase coherence index of |B| measured by Cluster-1 (red)and ACE (blue). Letters a, b and c indicate scales τ = 10, 100 and 1000s, respectively. The bars indicate the inertial subrange of each spacecraftobtained from Figure 3.6. The inverse of the ion cyclotron frequency fci ∼0.12 Hz in the solar wind frame is τ ∼ 8.3 sec, which is near the peakregions of kurtosis and phase coherence index.
SOURCE: Chian and Miranda (2009)
(PCS) in which the phases of the Fourier modes are made completely equal. The
power spectrum of three data sets ORG, PRS and PCS are kept the same, but their
phase spectra are different. An average of over 100 realizations of the phase shuffling
is performed to generate the phase-randomized surrogate data set SPRS(τ). Cφ(τ)=0
indicates that the phases of the scales τ of the original data are completely random
(i.e., null phase synchronization), whereas Cφ(τ)=1 indicates that the phases are
fully correlated (i.e., total phase synchronization). Figure 3.12 displays the computed
variation of Cφ with the time scale τ for magnetic field fluctuations of Cluster and
ACE, whose behaviors follow that of kurtosis.
The upper panel of Figure 3.12 shows the variation of kurtosis as a function of the
time scale τ . For large scales (τ>∼ 103 sec) kurtosis is nearly zero, implying that the
magnetic field fluctuations are near-Gaussian (phase incoherent). For 10 sec<∼ τ
<∼103 sec, kurtosis increases as the time scale decreases which characterizes intermit-
71
tency and non-Gaussianity related to nonlinear energy cascade within the inertial
subrange seen in Figure 3.6. The lower panel of Figure 3.12 shows the variation of
the phase coherence index with τ which presents similar characteristics of kurtosis,
indicating that phase synchronization due to nonlinear multi-scale interactions is
responsible for intermittency. The inertial subranges for Cluster and ACE (obtained
from Figure 3.6) are marked with a bar in Figure 3.12. The results of Figure 3.12
display similar trend as the upstream results of Koga et al. (2007) obtained by the
Geotail data at the Earth’s bow shock in the sense that, as the scale τ decreases,
both kurtosis and phase coherence index increase until a certain scale where they
reach their respective maxima, and then both kurtosis and phase coherence index
start to decrease as τ decreases.
We conclude from Figure 3.12 that either kurtosis or phase coherence index can be
used to determine the degree of intermittency and phase synchronization in solar
wind turbulence. Both nonlinear techniques prove that the solar wind magnetic
field fluctuations, measured by Cluster and ACE, are intermittent (non-Gaussian)
exhibiting high degree of intermittency (non-Gaussianity) at small scales and low
degree of intermittency (near-Gaussianity) at large scales, in complete agreement
with Figures 3.8-3.11. It is interesting to point out that the period of 10 s of Alfven
waves analyzed by Eastwood et al. (2003) in the same Cluster event, from 04:02:30
UT to 04:10:00 UT (outside of our interval), is close to the peak of Figure 3.12 where
the intermittency is strongest. In addition, solar wind turbulence can be decomposed
into coherent (non-Gaussian) and incoherent (Gaussian) component using the local
intermittent measure analysis (ALEXANDROVA et al., 2008).
3.1.2 ICME event
In this subsection we study the scale dependence of kurtosis and phase coherence in
intermittent magnetic field turbulence measured at three different locations of the
solar-terrestrial environment: (1) in the solar photosphere, (2) in the solar wind, and
(3) on the ground. We perform a comparison of two scenarios: a non-ICME event in
February 2002 and an ICME event in January 2005.
72
0
2
4
6
8
Kur
tosi
s
AR 09802Quiet region
100
101
102
r (× 1.45 Mm)
0
0.5
1
Phas
e C
oher
ence
Ind
ex
(a) (b)
SOHO MDI solar image at 22:24 UT on 1 February 2002
FIGURE 3.13 - (a) SOHO MDI solar image taken at 22:24 UT on 1 February 2002. Thetwo white squares enclose two areas containing the active region AR 09802(upper), and a quiet region (lower). (b) Kurtosis (upper panel) and thephase coherence index (lower panel) as a function of spatial scale r com-puted from AR 09802 and the quiet region.
SOURCE: Miranda et al. (2009)
0
2
4
6
8
Kur
tosi
s
AR 10720Quiet region
100
101
102
r (× 1.45 Mm)
0
0.5
1
Phas
e C
oher
ence
Ind
ex
SOHO MDI solar image at 04:47 UT on 16 January 2005
(a) (b)
FIGURE 3.14 - (a) SOHO MDI solar image taken at 04:47 UT on 16 January 2005. Thetwo white squares enclose two areas containing the active region AR 10720(upper), and a quiet region (lower). (b) Kurtosis (upper panel) and thephase coherence index (lower panel) as a function of spatial scale r com-puted from AR 10720 and the quiet region.
SOURCE: Miranda et al. (2009)
73
3.1.2.1 Non-Gaussianity and phase synchronization in the solar-
terrestrial environment
Solar images obtained by the SOHO MDI instrument provide the measurement of
the magnetic field in the solar photosphere. Near the centre of the solar image the
projection effects are negligible, hence the MDI solar image represents the vertical
or line-of-sight component of the photospheric magnetic field B‖ (ABRAMENKO et
al., 2002). Figure 3.13(a) shows a solar magnetogram taken by SOHO MDI on 1
February 2002. The white color corresponds to positive magnetic polarity, and the
black color corresponds to negative magnetic polarity. The two white squares mark
two selected areas, one containing the active region AR 09802 (upper) and the
other containing a quiet region (lower). Figure 3.13(b) shows kurtosis and the phase
coherence index as a function of spatial scale r computed from the two selected
regions. From Figure 3.13(b) we observe that the variation of kurtosis with r in the
quiet region is close to a Gaussian process (K = 0), being scale-invariant for all scales
r>∼2 pixels ∼ 2.9 megametres (Mm), consistent with the features of a monofractal
process (ABRAMENKO et al., 2002). The active region, on the other hand, displays
an increase of kurtosis as the spatial scale r decreases, which is a characteristic
of a non-Gaussian process related to nonlinear energy cascade within the inertial
subrange, and these values are higher than those obtained from the quiet region for
scales<∼20 pixels ∼ 29 Mm. The degree of phase synchronization measured by the
phase coherence index (HADA et al., 2003; KOGA et al., 2007; CHIAN; MIRANDA, 2009)
in the active region increases with decreasing spatial scale r, while the quiet region
presents low-degree of synchronization at all scales. Note that at large scales the
lack of datapoints introduces big errors in the computation of the phase coherence
index.
Figure 3.14(a) shows a SOHO MDI solar magnetogram obtained on 16 January 2005.
The two white squares enclose two selected areas containing the active region AR
10720 (upper) and a quiet region (lower), respectively. On 19-20 January 2005 several
flares associated with CMEs were observed in AR 10720, however in this period the
active region is too close to the solar limb, and the projection effects cannot be
neglected. Hence, we restrict our analysis to this earlier solar image when AR 10720
is near the disk centre. Figure 3.14(b) shows kurtosis and the phase coherence index
as a function of spatial scale r. It shows that the kurtosis of AR 10720 increases as
the scale r decreases, while kurtosis of the quiet region displays scale-invariance for
74
all scales r>∼2 pixels ∼ 2.9 Mm, similar to Fig. 3.13(b). Likewise, the phase coherence
index in Fig. 3.14(b) presents similar behavior as Fig. 3.13(b).
Now we direct our attention to the interplanetary magnetic field data collected
in situ in the solar wind. Figure 3.15(a) shows the time series of the modulus of
magnetic field |B| obtained by ACE and Cluster from 19:40:40 UT on 1 February
2002 to 03:56:38 UT on 3 February 2002. During this interval Cluster is in the
solar wind upstream of the Earth’s bow shock (CHIAN; MIRANDA, 2009). Although
Fig. 3.13(a) indicates the presence of several solar active regions on 1 February
2002, no M- or X-class solar flares occurred during the selected interval, and strong
interplanetary disturbances such as ICMEs were not seen. The upper panel of Figure
3.15(b) shows the variation of kurtosis as a function of time scale τ for magnetic field
fluctuations of ACE and Cluster. For 10 s<∼ τ
<∼ 103 s, kurtosis increases as the time
scale decreases, which characterizes non-Gaussianity. The lower panel of Fig. 3.15(b)
shows the variation of the phase coherence index with τ which presents a behavior
similar to kurtosis. From Fig. 3.15(b) we observe that the behavior of kurtosis and
the phase coherence index detected by ACE and Cluster are very similar except for
scales around 10 s, where ACE observes a higher level of intermittency and phase
synchronization than Cluster.
Figure 3.16(a) shows the time series of the modulus of magnetic field |B| measured
by ACE for the ICME event of 21-22 January 2005 (FOULLON et al., 2007). We se-
lected two intervals from this event. The first interval is located upstream of the
ICME shock which begins at 06:00:00 UT on 21 January and ends at 16:00:00 UT
on 21 January. The second interval is located downstream of the ICME shock which
begins at 16:47:19 UT on 21 January, and ends at 21:20:00 UT on 22 January.
Arrows indicate the selected intervals in Figure 3.16(a). In order to ensure the sta-
tionarity of data we avoid the “foot” transition region associated with the ICME
shock front. Figure 3.16(b) shows kurtosis and the phase coherence index as a func-
tion of time scale τ computed from the modulus of magnetic field |B| in upstream
and downstream regions of the ICME shock. It is evident, from Figure 3.16(b), that
both upstream and downstream regions indicate the features of intermittency and
phase synchronization across scales; moreover, the level of intermittency and phase
synchronization are higher in the downstream region than the upstream region for
all scales τ .
We analyze next the Earth’s geomagnetic field data obtained from two ground mag-
75
0
10
20
0
10
20
19:40:40Day 32
11:48:39Day 33
03:56:38Day 34
0
50
100ACE
Cluster
| |
| |
100
101
102
103
104
0
0.5
1
ACE(a)
|B| (
nT
)|B
| (nT
)
Time (UT)
Cluster
τ (sec)
(b)
Ph
ase
Co
her
ence
In
dex
Ku
rto
sis
Solar wind event of 1−3 February 2002
FIGURE 3.15 - (a) Time series of the modulus of magnetic field |B| measured by ACE(upper panel) and Cluster (lower panel) for the solar wind event of 1-3February 2002. (b) Kurtosis (upper panel) and the phase coherence index(lower panel) of |B| as a function of time scale τ . Horizontal bars indicatethe inertial subranges of ACE and Cluster, respectively.
SOURCE: Miranda et al. (2009)
| |
0
50
100Downstream
Upstream
100
101
102
103
104
0
0.5
1
0
20
40
06:00:00Day 21
16:00:00Day 21
16:47:19Day 21
21:20:00Day 22
τ (sec)
ACE (b)
ICME event of 21−22 January 2005
|B| (
nT
)
(a)
Time (UT)
Ku
rto
sis
Ph
ase
Co
her
ence
In
dex
FIGURE 3.16 - (a) Time series of the modulus of magnetic field |B| measured by ACE forthe ICME event of 21-22 January 2005. (b) Kurtosis (upper panel) andthe phase coherence index (lower panel) of |B| as a function of time scaleτ for the upstream and downstream regions of the ICME shock.
SOURCE: Miranda et al. (2009)
76
0
5
10
15
20
|B| (
nT)
-200
0
200
400
|B| (
nT)
-2
0
2
Pc3
-2
0
2
Pc4
-20
0
20
Pc5
0
50
100
Kur
tosi
s
ACEJPA
100
101
102
103
104
0
0.5
1
Pha
se C
oher
ence
Inde
x
Ground magnetometer data at Ji-Parana (JPA), Brazil(Geomagnetic latitude: ~ 0
o)
Time (UT)
19:40:40Day 32
11:48:39Day 33
03:56:38Day 34
τ (sec)
ACE
JPA
´
FIGURE 3.17 - (a) From top to bottom: time series of |B| measured by ACE for the solarwind event of 1-3 February 2002; modulus of the Earth’s geomagnetic field|B| measured by a ground magnetometer at Ji-Parana, Brazil, during thesame time interval; time series of Pc3 (10-45 s), Pc4 (45-150 s) and Pc5(150-1000 s) micropulsations. (b) Kurtosis (upper panel) and the phasecoherence index (lower panel) of |B| measured by ACE and the groundmagnetometer at JPA as a function of time scale τ.
SOURCE: Miranda et al. (2009)
0
20
40
|B| (
nT)
-200
0
200
|B| (
nT)
-15
0
15
Pc3
-15
0
15
Pc4
-15
0
15
Pc5
06:00:00Day 21
21:20:00Day 22
16:00:00Day 21
16:47:19Day 21
| |
0
10
20
Kur
tosi
s
VSS DownstreamVSS Upstream
100
101
102
103
104
0
0.5
1
Phas
e C
oher
ence
Ind
ex
Ground magnetometer data at Vassouras (VSS), Brazil(Geomagnetic latitude: ~ 19
o)
Time (UT) τ (sec)
ACE
VSS
FIGURE 3.18 - (a) From top to bottom: time series of |B| (nT) measured by ACE for theICME event of 21-22 January 2005; modulus of the Earth’s geomagneticfield |B| (nT) measured by a ground magnetometer at Vassouras, Brazil;time series of Pc3 (10-45 s), Pc4 (45-150 s) and Pc5 (150-1000 s) micropul-sations. (b) Kurtosis (upper panel) and the phase coherence index (lowerpanel) of |B| measured by the ground magnetometer at VSS as a functionof time scale τ .
SOURCE: Miranda et al. (2009)
77
netometers. The top panel of Fig. 3.17(a) shows the time series of the modulus of
magnetic field |B| observed by ACE at L1 during the non-ICME solar wind event of
1-3 February 2002. The second panel shows the modulus of the Earth’s geomagnetic
field measured by a ground magnetometer at Ji-Parana (JPA), Brazil (geomagnetic
latitude ∼ 0o). The three bottom panels show the time series of Pc3 (10-45 s), Pc4
(45-150 s) and Pc5 (150-1000 s) continuous geomagnetic pulsations, respectively.
Each time series of geomagnetic micropulsations is obtained by applying a Fourier
band-pass filter. Fig. 3.17(b) shows kurtosis and the phase coherence index of ACE
magnetic field data and JPA geomagnetic field data as a function of time scale τ . Fig.
3.17(b) indicates that the geomagnetic field fluctuations measured on the ground are
intermittent. For almost all scales the level of intermittency and phase coherence of
ACE are higher than the JPA.
Finally, for the ICME event of 21-22 January 2005, we plot in the top panel of
Figure 3.18(a) the time series of the modulus of magnetic field |B| observed by ACE
and in the second panel the modulus of the Earth’s geomagnetic field measured by
a ground magnetometer at Vassouras (VSS), Brazil (geomagnetic latitude ∼ 19o).
The three bottom panels show the time series of Pc3, Pc4 and Pc5 geomagnetic
micropulsations. All the time series of VSS were shifted by −1462 s to synchronize
with the ICME shock arrival at ACE. After shifting, we divide the VSS geomagnetic
field time series into“upstream”and“downstream”intervals in analogy with the ACE
magnetic field data of the ICME shock. Figure 3.18(b) shows kurtosis and the phase
coherence index of the VSS upstream and downstream intervals as a function of time
scale τ . It shows that for all scales the level of intermittency and phase coherence are
higher in the “downstream” geomagnetic field-fluctuations after the arrival of ICME
than the “upstream” geomagnetic field fluctuations before the arrival of ICME.
A comparison of the variation of kurtosis and the phase coherence index as a function
of τ for the downstream intervals observed by ACE in the solar wind (Fig. 3.16(b))
and the VSS ground magnetometer (Fig. 3.18(b)) reveals a common feature consist-
ing of three peaks between scales τ ∼ 304 s and τ ∼ 1860 s, marked with a bar in
both figures. This can be interpreted as evidence of a close correlation of the Earth’s
geomagnetic field with the ICME driver at these scales.
78
3.2 Synchronization in atmospheric turbulence
The aim of this section is to apply the kurtosis (fourth-order structure function)
and phase coherence techniques to determine the intermittent nature of day-time
atmospheric turbulence above and within the Amazon forest canopy. In particular,
we show that both techniques are capable of characterizing the dissimilarity of scalar
and velocity in above-canopy and in-canopy atmospheric turbulence.
3.2.1 Amazon forest canopy data
A major atmospheric mesoscale campaign during the wet season of the Large-Scale
Biosphere-Atmosphere Experiment in Amazonia (LBA) was carried out in January-
March 1999 (SILVA DIAS et al., 2002). LBA is an international project led by Brazil
designed to study: 1) the climatological, ecological, biogeochemical, and hydrological
functions of the Amazon region, 2) the impact of land uses caused by deforestation,
and 3) the interactions between Amazonia and the Earth system. As part of this
campaign, a micrometeorological tower was set up in a southwest Amazon basin at
the Biological Reserve of Jaru (Rebio Jaru: 1004’S, 6156’W) in Rondonia State
of Brazil; with instruments located at two different heights: above-canopy at 66 m
and inside-canopy at 21 m, to make simultaneous measurements of eddy covari-
ance and vertical profiles of air temperature, wind velocity, radiation and humidity.
Three-dimensional wind velocity and air temperature measurements were made at
a sampling rate of 60 Hz, using sonic anemometers and thermometers.
In this section we investigate the atmospheric turbulence data taken from noon to
12:30 on Julian Day 068, when the forest crown is heated by the solar radiation.
A dataset of 108000 points for temperature and vertical velocity is used for this
analysis. During this period the top of the canopy is hotter than its surroundings,
thus temperatures decrease both upwards towards the atmosphere above the canopy
and downwards towards the ground surface, resulting in an unstably stratified atmo-
sphere above the canopy and a stably stratification region inside the canopy. Figure
3.19 shows the original time series of temperature T and vertical wind velocity w
above and within the Amazon forest canopy.
3.2.2 Intermittency and phase coherence in atmospheric turbulence
Figures 3.20 and 3.21 show, respectively, the scale dependence of the normalized
temperature-difference ∆T = (δT− < δT >)/σT and the normalized vertical wind
79
12:00:00 12:15:00 12:30:00-2
-1
0
1
2
12:00:00 12:15:00 12:30:0028
28.2
28.4
28.6
28.8
29
12:00:00 12:15:00 12:30:0028
28.2
28.4
28.6
28.8
29
12:00:00 12:15:00 12:30:00-2
-1
0
1
2
T (
C)
o
ABOVE−CANOPY IN−CANOPY
Local Time
Local Time Local Time
Local Time
T (
C)
o
w (
m/s
)
w (
m/s
)
FIGURE 3.19 - Time series of temperature T and vertical wind velocity w above theAmazon forest canopy (left panels) and within the Amazon canopy (rightpanels), taken from noon to 12:30 on Julian Day 068, 1999.
SOURCE: Chian et al. (2008)
velocity-difference ∆w = (δw− < δw >)/σw above and within the Amazon forest
canopy, respectively, for three different time scales (τ = 0.15 s, 3.3 s and 100.0 s),
where δu = u(t+ τ)−u(t) denotes the two-point difference of vertical wind velocity
w or temperature T for a given time scale τ , the brackets denote the mean value
of δu, and σu denotes the standard deviation of δu. Evidently, the fluctuations in
Figures 3.20 and 3.21 become more intermittent as the scale becomes smaller.
The intermittent characteristics of Amazon atmospheric turbulence can be elu-
cidated by the probability density function (PDF) of the temperature and ver-
tical wind velocity fluctuations. Figure 3.22 shows the PDF of the normalized
temperature-difference ∆T (right panels) and the normalized vertical wind velocity-
difference ∆w (left panels) fluctuations above the Amazon forest canopy for 3 dif-
ferent scales (τ = 0.15 s, 3.3 s and 100.0 s, indicated by a, b and c, respectively,
in Figure 3.25), superposed by a Gaussian PDF (gray line). Figure 3.23 shows the
corresponding PDFs inside the Amazon forest canopy. It follows, from Figures 3.22
and 3.23, that for both temperature and vertical wind velocity the PDF is close to
the Gaussian distribution at large scale (τ=100 s) but deviates substantially from
the Gaussian distribution as the scale τ decreases. At small scales (τ=0.15 s and
80
0 200 400 600 800
0 200 400 600 800
T∆
w∆
τ = 100 s
τ = 0.15 s
τ = 3.3 s
τ = 0.15 s
τ = 3.3 s
τ = 100s
Time (s)
Time (s)
ABOVE−CANOPY
FIGURE 3.20 - Scale dependence of the normalized temperature-difference ∆T and thenormalized vertical wind velocity-difference ∆w above the Amazon forestcanopy for three different time scales (τ = 0.15 s, 3.3 s and 100 s).
SOURCE: Chian et al. (2008)
3.3 s) the shape of PDF becomes leptokurtic, exhibiting fat tails and sharp peaks.
Figures 3.22 and 3.23 show that the tails of PDF get longer and the peaks of PDF
get sharper as the scale decreases.
Intermittency can be quantified by calculating the variation of the normalized fourth-
order structure function K (kurtosis) with scale τ , K(τ) = (1/n)∑n
i=1((δui− <
δui >)/σu)4 − 3. For an intermittent signal K(τ) > 0 and K increases when scale
decreases; for a Gaussian noise K=0 for all scales. Top panels of Figures 3.24 and
3.25 show the computed variation of K with scale τ for temperature and vertical
wind velocity, above and inside the Amazon forest canopy, where the horizontal bar
denotes the inertial subrange, approximately from 0.15 s to 3.3 s (BOLZAN et al.,
2002; RAMOS et al., 2004), estimated by the method of Kulkarni et al. (1999) using
isotropy coefficient calculated by the Haar wavelet transform.
An alternative way of quantifying intermittency is to apply the phase coherence
81
0 200 400 600 800
0 200 400 600 800
T∆
w∆
τ = 0.15
τ = 3.3
s
s
τ = 3.3 s
τ = 0.15 s
τ = 100s
τ = 100s
IN−CANOPY
Time (s)
Time (s)
FIGURE 3.21 - Scale dependence of the normalized temperature-difference ∆T and thenormalized vertical wind velocity-difference ∆w within the Amazon forestcanopy for three different time scales (τ = 0.15 s, 3.3 s and 100 s).
SOURCE: Chian et al. (2008)
technique using surrogate data by defining a phase coherence index (HADA et al.,
2003; KOGA et al., 2007; KOGA et al., 2008), Cφ(τ) = (SPRS(τ)−SORG(τ))/(SPRS(τ)−SPCS(τ)). This index measures the degree of phase coherence in an original data set
(ORG) by comparing it with two surrogate data sets created from the original data
set: the phase-randomized surrogate (PRS) in which the phases of the Fourier modes
are made completely random, and the phase-correlated surrogate (PCS) in which the
phases of the Fourier modes are made completely equal. The power spectrum of three
data sets ORG, PRS and PCS are kept the same, but their phase distributions are
different. Each length of ORG, PRS and PCS data set is measured by the magnitude
of first-order structure function S1(τ) =∑n
i=1 |ui+τ − ui|. An average of over 100
realizations of the phase shuffling is performed to generate the phase-randomized
surrogate data set SPRS(τ). Cφ(τ)=0 indicates that the phases of the scales τ of the
original data are completely random, whereas Cφ(τ)=1 indicates that the phases are
completely correlated. Bottom panels of Figures 3.24 and 3.25 show the computed
82
-10 -5 0 5 1010
-5
10-4
10-3
10-2
10-1
100
-10 -5 0 5 1010
-5
10-4
10-3
10-2
10-1
100
-10 -5 0 5 1010
-5
10-4
10-3
10-2
10-1
100
-10 -5 0 5 1010
-5
10-4
10-3
10-2
10-1
100
-10 -5 0 5 1010
-5
10-4
10-3
10-2
10-1
100
-10 -5 0 5 1010
-5
10-4
10-3
10-2
10-1
100
ABOVE−CANOPY
PD
FP
DF
PD
F
s
s
sτ = 100
τ = 3.3
τ = 0.15
Vertical Velocity
w
w
w
∆
∆
∆
PD
FP
DF
PD
F
s
s
s
Temperature
τ = 0.15
τ = 3.3
τ = 100
T
T
T
∆
∆
∆
FIGURE 3.22 - Probability density distribution (PDF) of the normalized vertical windvelocity-difference ∆w (left panels) and the normalized temperature-difference ∆T (right panels) fluctuations above the Amazon forest canopyfor three different scales (τ = 0.15 s, 3.3 s and 100.0 s), superposed by aGaussian PDF (gray line).
SOURCE: Chian et al. (2008)
variation of Cφ with scale τ for temperature and vertical wind velocity above and
inside the Amazon forest canopy.
Figure 3.24 confirms that both kurtosis and phase coherence index can be used to
measure the degree of intermittency and phase coherence in turbulence. Left panels
of Figure 3.24 show that for all scales τ<∼ 35 s the in-canopy vertical velocity is more
intermittent than the above-canopy vertical velocity. Right panels of Figure 3.24
show that for small scales outside the inertial subrange the in-canopy temperature
is more intermittent than the above-canopy temperature; whereas, for scales within
and greater than the inertial subrange the in-canopy temperature is less intermittent
than the above-canopy temperature.
83
-10 -5 0 5 1010
-5
10-4
10-3
10-2
10-1
100
-10 -5 0 5 1010
-5
10-4
10-3
10-2
10-1
100
-10 -5 0 5 1010
-5
10-4
10-3
10-2
10-1
100
-10 -5 0 5 1010
-5
10-4
10-3
10-2
10-1
100
-10 -5 0 5 1010
-5
10-4
10-3
10-2
10-1
100
-10 -5 0 5 1010
-5
10-4
10-3
10-2
10-1
100
IN−CANOPY
PD
FP
DF
PD
F
s
s
s
τ = 0.15
τ = 3.3
τ = 100
∆w
w∆
w∆
PD
FP
DF
PD
F
s
s
s
τ = 0.15
τ = 3.3
τ = 100
T∆
T∆
T∆
Vertical Velocity Temperature
FIGURE 3.23 - Probability density distribution (PDF) of the normalized vertical windvelocity-difference ∆w (left panels) and the normalized temperature-difference ∆T (right panels) fluctuations within the Amazon forest canopyfor three different time scales (τ = 0.15 s, 3.3 s and 100.0 s), superposedby a Gaussian PDF (gray line).
SOURCE: Chian et al. (2008)
3.2.3 Scalar-velocity dissimilarity in atmospheric intermittent turbu-
lence
The dissimilarity between temperature (scalar) and vertical wind velocity (momen-
tum) can be elucidated by Figures 3.22, 3.23 and 3.24. Figure 3.22 shows that for
scales τ=0.15 s and 3.3 s (a and b in Figure 3.25) the PDFs of temperature fluc-
tuations have longer tails and sharper peaks than the PDFs of vertical velocity
fluctuations, implying that at these scales temperature is more intermittent than
vertical wind velocity above the Amazon forest canopy. The scalar-velocity dissimi-
larity above the canopy is clearly characterized by both kurtosis and phase coherence
techniques in the left panels of Figure 3.25, which show that the above-canopy tem-
perature is more intermittent than vertical wind velocity for all scales τ<∼ 55 s.
In contrast, the right panels of Figure 3.25 show that the scalar-velocity dissimilarity
of the in-canopy turbulence is greatly reduced. The scalar-velocity similarity inside
84
| |
0.01 0.1 1 10 100 10000
0.1
0.2
0.3
0.4
0.5
0.6Above-canopyIn-canopy
| |
0.01 0.1 1 10 100 10000
0.1
0.2
0.3
0.4
0.5
0.6Above-canopyIn-canopy
| |
0.01 0.1 1 10 100 10000
10
20
30
40 Above-canopyIn-canopy
| |
0.01 0.1 1 10 100 10000
10
20
30
40 Above-canopyIn-canopy
TemperatureVertical Velocity
Kur
tosi
s
τ τPh
ase
Coh
eren
ce I
ndex
ττ
Phas
e C
oher
ence
Ind
exK
urto
sis
(sec)
(sec) (sec)
(sec)
FIGURE 3.24 - Kurtosis and phase coherence index of vertical wind velocities (left pan-els) and temperatures (right panels) above and within the Amazon forestcanopy. Thick solid lines denote the in-canopy vertical wind velocity, thindash-dotted lines denote the above-canopy vertical wind velocity, thickdashed lines denote the in-canopy temperature, and thin dotted lines de-note the above-canopy temperature. The bar indicates the inertial sub-range (approximately from τ = 0.15 s to 3.3 s).
SOURCE: Chian et al. (2008)
the canopy is evidenced in Figure 3.23 which shows that, at τ=0.15 s the PDFs
of temperature and vertical wind velocity have similar sharp peaks and long-tail
statistics.
3.3 Discussion
For numerical simulations and analytic formulation of turbulence based on a set of
deterministic plasma or fluid equations, in the absence of noise, it is natural to ex-
pect that a departure from Gaussianity arises from nonlinear multiscale coupling in
turbulent energy cascade (FRISCH, 1995; DAVIDSON, 2004). In contrast, the observa-
tional data of space plasma turbulence is an admixture of deterministic signal and
stochastic noise. Under this circumstance, a demonstration of finite phase synchro-
nization is required to ascertain the nonlinear origin of non-Gaussian fluctuations.
For magnetic field turbulence in the solar wind, in the absence of strong disturbances
85
| |
0.01 0.1 1 10 100 10000
0.1
0.2
0.3
0.4
0.5
0.6Vertical velocityTemperature
| |
0.01 0.1 1 10 100 10000
0.1
0.2
0.3
0.4
0.5
0.6Vertical velocityTemperature
| |
0.01 0.1 1 10 100 10000
10
20
30
40 Vertical velocityTemperature
| |
0.01 0.1 1 10 100 10000
10
20
30
40 Vertical velocityTemperature
τ
Pha
se C
oher
ence
Inde
x
Pha
se C
oher
ence
Inde
xK
urto
sis
Kur
tosi
s
τ τ (sec)(sec)
τ(sec) (sec)
ABOVE−CANOPY IN−CANOPY
a
bc
a
cb
FIGURE 3.25 - Kurtosis and Phase Coherence Index of vertical wind velocities and tem-peratures above (left panels) and within (right panels) the Amazon forestcanopy. Thick solid lines denote the in-canopy vertical wind velocity, thindash-dotted lines denote the above-canopy vertical wind velocity, thickdashed lines denote the in-canopy temperature, and thin dotted lines de-note the above-canopy temperature. Letters a, b and c indicate scales τ= 0.15 s, 3.3 s and 100.0 s, respectively. The bar indicates the inertialsubrange (approximately from τ = 0.15 s to 3.3 s).
SOURCE: Chian et al. (2008)
such as ICMEs (subsection 3.1.1) we showed that, within the inertial subrange, the
departure from Gaussianity increases as the scale decreases, which is a characteristic
of intermittent turbulence. Moreover, Figure 3.12 shows that phase synchronization
associated with nonlinear multiscale interactions is the origin of intermittency and
non-Gaussianity, which leads to the formation of large-amplitude phase coherent
(intermittent) structures at small scales. We have identified these large-amplitude
phase coherent structures as spikes in the time series of magnetic-field differences
in Figure 3.24 and fat tails in the PDF and extended flatness in Figure 3.9. Our
computed results are consistent with the Helios analysis by Bruno et al. (2003) of
solar wind intermittency in the inner heliosphere. For slow solar wind at 0.9 AU,
Bruno et al. (2003) obtained the values of 19 for τ = 100 s and 7 for τ = 1000 s for
flatness, which are close to our numerical examples given in Table 3.1.
86
Large-amplitude coherent structures embedded within intermittent magnetic field
turbulence in the foreshock region of Earth’s bow shock have been detected by
Cluster. Eastwood et al. (2003) detected, on 3 February 2002, large-amplitude ultra-
low-frequency (ULF) Alfven waves with wavelength of 3,400 km, in the foreshock
region of Earth’s quasi-parallel bow shock. Stasiewicz et al. (2003) reported the mea-
surement, on 3 February 2002, of the density profiles and wave spectra inside fast
magnetosonic shocklets, 1,000 km in size and amplitude of 10 times the ambient
magnetic field, upstream of a quasi-parallel bow shock. Behlke et al. (2004) ob-
served on 3 February 2002 solitary waves, as bipolar pulses in the spiky electric field
moving at velocities of 400-1,200 km/s along the ambient magnetic field with peak-
to-peak amplitudes of E‖ = 65 mV/m and parallel scale sizes of L‖ ∼ 300− 600 m
∼ 10λD (Debye length), within short large-amplitude magnetic structures (SLAMS)
upstream of a quasi-parallel bow shock. In addition, Parks et al. (2006) detected
ion density holes accompanied by magnetic holes (∼ 3,700 km) upstream of a quasi-
parallel bow shock, which are seen only with upstream particles, suggesting a link
with backstreaming particles interacting with the solar wind. All the aforementioned
observations (found outside of the time interval analyzed in this subsection) refer
to the foreshock region of a quasi-parallel shock which is a patchwork of SLAMS
slowing down and piling up. SLAMS evolve from ULF instabilities excited by coun-
terstreaming plasma populations (SCHWARTZ, 2006) and have amplitudes 2-4 times
larger than the ambient magnetic field, with typical durations of around 10 s and
transverse dimensions of ∼ 1 RE (LUCEK et al., 2004). Although the interval of Clus-
ter data analyzed in this subsection is outside of the foreshock region, we expect that
some of the coherent structures discussed above may contribute to the intermittent
turbulence in this upstream region when IMF connects to bow shock (KELLOGG;
HORBURY, 2005).
Large-amplitude coherent structures in the ambient solar wind have been found by
Helios and ACE. Bruno et al. (2001), Bruno et al. (2003) and Bruno et al. (2005)
reported the observation of coherent structures in the intermittent magnetic field
turbulence between 0.3 to 1 AU using the Helios solar wind data. During a fast solar
wind interval on Julian Day 49-52 1976 when Helios-2 was at 0.9 AU and the solar
wind fluctuations are Alfvenic, they detected one coherent structure related to a flux
tube of scale size ∼ 9 × 105 km (BRUNO et al., 2001). In a recent paper, Borovsky
(2008) studied the statistics of 65,860 flux tubes in the ACE data for the period
1998-2004, and obtained a median scale size of flux tubes of ∼ 6.2× 105 km for slow
87
solar wind and of ∼ 4.2×105 km for fast solar wind. These coherent structures (flux
tubes) are spotted by large changes in the magnetic field direction and the vector
flow velocity, and are associated with large changes in the ion entropy density and
the alpha-to-proton ratio. These flux tubes map to granule and supergranule sizes
on the Sun’s photosphere. Borovsky (2008) suggested a method for using these solar
wind coherent structures for remote sensing of the dynamics of the Sun’s magnetic
carpet. Note that the coherent structures reported by Bruno et al. (2001), Bruno
et al. (2003), Bruno et al. (2005) and Borovsky (2008) have spatial scales ∼ 105
km which correspond to time scales ∼ 103 s (assuming Taylor’s hypothesis and
Vsw ∼ 4 × 102 km/s). These large-scale flux tubes might be coherent structures
convected by the solar wind from the base of the solar atmosphere (BRUNO et al.,
2001; BOROVSKY, 2008).
It is worth mentioning that the most intermittent events in the solar wind turbu-
lence, which occur at time scales of the order of a few minutes, have been identi-
fied as current sheets and shock waves (VELTRI; MANGENEY, 1999; VELTRI, 1999;
ALEXANDROVA et al., 2008). These small-scale intermittent structures may be related
to locally occurring magnetic reconnections (CHANG et al., 2004; ZHOU et al., 2004;
GRECO et al., 2009) and may play an important role in the phase synchronization
observed by Figure 3.12 in the interplanetary magnetic field turbulence.
In section 3.1.2 we studied and compared non-Gaussianity and phase synchroniza-
tion in solar-terrestrial magnetic field turbulence in two scenarios: a non-ICME event
and an ICME event. For each scenario, we calculated first kurtosis and the phase
coherence index for an active region and a quiet region, respectively, in the solar
photosphere. We then applied the same techniques to solar wind turbulence using
in situ data. The response of the Earth’s geomagnetic field using data from ground
magnetometers in Brazil was presented. In particular, we showed that the interplan-
etary magnetic field turbulence downstream of the ICME shock is closely correlated
with the geomagnetic turbulence detected on the ground.
In section 3.2 we analyzed day-time atmospheric data above and within an Amazon
forest canopy. We showed that the scale dependence of kurtosis and phase coherence
index for vertical wind velocity and temperature above and within canopy exhibit
similar behaviors, as seen in Fig. 3.24. In particular, both techniques demonstrate
a clear enhancement of scalar-velocity similarity for in-canopy turbulence in com-
parison to its above-canopy counterpart, as seen in Figure 3.25. Our results prove
88
that the atmospheric intermittent turbulence, above and within the Amazon forest
canopy, is generated by the phase coherence due to nonlinear wave-wave interactions.
Turbulence consists of an admixture of chaos and noise. Recent studies have identi-
fied the chaotic nature of the solar-terrestrial environment (MACEK, 1998; CHIAN et
al., 2006) and the atmospheric turbulence above the Amazon forest (CAMPANHARO
et al., 2008), which have been confirmed by computer simulations of temporal chaos
and spatiotemporal chaos in fluids and plasmas (CHIAN et al., 2006; REMPEL; CHIAN,
2007; REMPEL et al., 2007). It is likely that phase coherence and chaotic synchro-
nization are the origin of energy bursts and coherent structures of turbulence in the
complex earth-ocean-space system (HE; CHIAN, 2003; HE; CHIAN, 2005).
The results of this chapter indicate that variations of kurtosis and the phase co-
herence index are similar across scales. Since kurtosis can be interpreted as the
energy squared (see subsection 2.1.3), it can also be regarded as a measurement
of non-Gaussianity (i.e., synchronization) using the amplitude information. Hence,
our observational results show that there is a duality between phase synchronization
(quantified by the phase coherence index) and amplitude synchronization (quantified
by kurtosis).
89
4 THEORY OF SYNCHRONIZATION IN SPATIOTEMPORAL IN-
TERMITTENCY
In this Chapter the mathematical tools described in Section 2.2 are applied to the
Benjamin-Bona-Mahony (BBM) equation, a nonlinear model of drift waves in mag-
netized plasmas and shallow water waves in fluids. This model is also called the
regularized long-wave equation (RLWE). In Section 4.1 we present the equation and
demonstrate that the coupling of chaotic saddles is responsible for the on-off inter-
mittency at the onset of spatiotemporal chaos. Section 4.2 is devoted to measure the
degree of amplitude-phase synchronization in the BBM equation. We perform an
analysis of the amplitude and phase synchronization due to multiscale interactions
in the temporally and spatiotemporally chaotic attractor embedded in the chaotic
attractor at the onset of spatiotemporal chaos.
4.1 Spatiotemporal intermittency and chaotic saddles in the Benjamin-
Bona-Mahony (BBM) equation
4.1.1 The Benjamin-Bona-Mahony equation
The unidirectional propagation of long waves in fluids with small but finite amplitude
in systems with nonlinearity and dispersion can be described by the Korteweg-de
Vries (KdV) equation, which in dimensionless form is given by (BENJAMIN et al.,
1972; DODD et al., 1982; NICHOLSON, 1983)
∂tu+ ∂xu+ u∂xu+ ∂xxxu = 0. (4.1)
Equation (4.1) is often written without the second term, after taking x′ = x − t
and t as independent variables. The KdV equation was originally derived for water
waves and has been used as a model for long waves in many other physical systems
(DODD et al., 1982), e.g., nonlinear ion-acoustic waves in plasmas (NICHOLSON, 1983).
However, some mathematical properties of the KdV equation such as existence and
stability of solutions are difficult to study (BENJAMIN et al., 1972). Furthermore,
there are other problems, mainly due to the dispersion term. For example, from the
linear dispersion relation of Eq. (4.1)
ω = k − k3, (4.2)
91
one can observe that the phase velocity Vph = ω/k becomes negative for k2 > 1,
which is inconsistent with the assumption of forward-travelling waves, used in the
derivation of Eq. (4.1) (BENJAMIN et al., 1972). Moreover, the group velocity
dω
dk= 1− 3k2, (4.3)
does not have any lower limit, thus there is no bound to the velocity with which
short wavelength features propagate. These impediments have led to the pursuit of
an alternative model. The regularized long-wave equation was proposed by Peregrine
(1966) and Benjamin et al. (1972) as an alternative to the KdV equation. It was
later derived by He and Salat (1989) as a model for nonlinear drift waves in plasmas,
with a periodic driving term and a linear damping term introduced ad hoc in order
to study transition to chaos. The driven-damped regularized long wave equation is
given by (HE; SALAT, 1989; REMPEL; CHIAN, 2007; HE, 1998; HE; CHIAN, 2004; HE;
CHIAN, 2003; HE; CHIAN, 2005)
∂tu+ c∂xu+ fu∂xu+ a∂txxu = −νu− ε sin(κx− Ωt), (4.4)
where a, c, and f are constants, ν is a damping parameter, ε is the driver amplitude,
κ is the driver wavenumber and Ω is the driver frequency. Originally, Peregrine
(1966) obtained Eq. (4.4) (without the two terms on the right hand side) from
the momentum equation for the mean horizontal velocity of water u(x, t) of an
irrotational flow by assuming that waves only travel in one direction and the ratios
between wave amplitude and water depth and between water depth and wavelength
are small, where wavelength means the distance in which significant changes in
surface height occur. The third-order derivative term in Eq. (4.4) expresses the effect
of the vertical acceleration of water on pressure. In the absence of forcing (ε = 0),
the linear dispersion relation is
ω(k) = (iν − ck)/(ak2 − 1), (4.5)
from which the phase velocity can be obtained
92
Vph = − c
ak2 − 1+
iν
k(ak2 − 1). (4.6)
Assuming a < 0 and c > 0, Re[Vph] > 0 for all k. The group velocity
Vg = − c
ak2 − 1+ 2ack2
(1
ak2 − 1
)2
− 2iaνk
(1
ak2 − 1
)2
(4.7)
approaches zero for large k, which means that high wavenumber features do not
propagate. Existence, uniqueness and stability of solutions of the RLWE have been
formally demonstrated by Benjamin et al. (1972).
We define periodic boundary conditions u(x, t) = u(x + 2π, t), and fix a = −0.287,
c = 1, f = −6, ν = 0.1, κ = 1 and Ω = 0.65. These parameter values can be chosen
arbitrarily, with the exception of a, which must be negative for physical reasons
and to avoid numerical instability (HE; SALAT, 1989). Here, the values are chosen
in order to study the transition to spatiotemporal chaos previously identified by He
(1998). Thus, the driver amplitude ε is the only control parameter. From Eq. (4.5),
since a < 0 and ν > 0, Im[ω(k)] < 0 for all k. Thus, all modes are linearly damped.
This is in contrast to the Kuramoto-Sivashinsky equation studied by Rempel et al.
(2007), where there was a band of linearly unstable Fourier modes. In the case of the
RLWE, the external driver is necessary to destabilize the mode with wave number
k = 1, as discussed below.
We solve equation (4.4) with the spectral (Galerkin) method, by expanding u(x, t)
in a Fourier series following Rempel and Chian (2007)
u(x, t) =N∑
k=−N
uk(t)eikx, (4.8)
where k = 2πn/L, n = −N, ..., N , L = 2π is the system length, and i =√−1. We
set N = 32. By introducing (4.8) into (4.4) one obtains a set of ordinary differential
equations in terms of the complex Fourier coefficients uk(t)
93
(1− ak2)dukdt
= −ickuk − νuk +ε
2sin(Ωt)δ1,k +
iε
2cos(Ωt)δ1,k − [ifkukuk]k, (4.9)
where δ is the Kronecker delta and the last term on the right-hand side is the
Fourier transform of fu∂xu, which is responsible for nonlinear wave coupling. In
order to compute this term, the pseudospectral method is used, where the derivative
is obtained in the Fourier space, ∂xu → ikuk. Next, both ikuk and uk are inverse-
Fourier transformed to real space, where the multiplication fu∂xu is performed.
Finally, the result is Fourier transformed again and inserted into equation (4.9).
Numerical integration is performed using the lsodar integrator (PETZOLD, 1983), a
variable-step integrator available in www.netlib.org. At each time step, 1/3 of the
high k modes are set to zero in order to avoid aliasing errors (TAJIMA, 1989). Thus,
the effective number of modes is N = 20. From equation (4.9), it can be seen that
in the presence of an external driver (ε 6= 0) energy is injected into the mode k = 1
and spreads towards other modes through the nonlinear term. Thus, the forcing and
nonlinear terms are responsible for driving the system away from equilibrium and
leading the system to chaos.
4.1.2 Transition to spatiotemporal chaos
The transition to spatiotemporal chaos (STC) can be easily recognized as a sudden
change in the spatiotemporal patterns of the numerical solutions of equation (4.4).
Figure 4.1 shows the asymptotic solutions obtained for two values of the control
parameter ε. The first regime (Fig. 4.1, upper panel), obtained for ε = 0.199, is
regular in space, but a positive value of the maximum Lyapunov exponent (λ1 ≈0.05) for the attractor indicates that the dynamics is temporally chaotic (TC). The
Lyapunov exponents are obtained from the eigenvalues of the linearized vector field
along a trajectory, and are computed with the code given in Wolf et al. (1985),
where a Gram-Schmidt orthonormalization process which employs the Euclidean
norm is used and base 2 logarithm is adopted for the exponents. The lower panel
of Fig. 4.1 shows that in the second regime (ε = 0.201) the spatiotemporal patterns
are disordered in both space and time. The maximum Lyapunov exponent for the
attractor STCA has jumped to λ1 ≈ 0.12. The transition from spatial regularity
to spatial irregularity is due to an attractor-widening crisis that occurs after the
collision of the spatially regular attractor with an unstable saddle orbit, and has
94
(x, t)u
020
60
020
4060
80
−10
10
0
40
tx/
(x, t)u
020
60
020
4060
80
−10
10
0
40
tx/
x∆
x∆
(ε = 0.199)TCA
(ε = 0.201)STCA
FIGURE 4.1 - Spatiotemporal patterns u(x, t) of the regularized long wave equation, forε = 0.199 (temporally chaotic attractor, TCA, upper panel), and ε = 0.201(spatiotemporally chaotic attractor, STCA, lower panel).
SOURCE: Rempel et al. (2009)
95
uu
FIGURE 4.2 - Time-averaged power spectra in the k wavenumber domain, for ε = 0.199(TCA regime, dashed line) and ε = 0.201 (STCA regime, solid line), forsimulations with N = 32 (left) and N = 512 (right) Fourier modes.
SOURCE: Rempel et al. (2009)
been described in a series of papers (HE, 1998; HE; CHIAN, 2004; HE; CHIAN, 2003).
The energy distribution among Fourier modes can be seen in the time-averaged
power spectra 〈|uk|2〉 depicted in Fig. 4.2 for N = 32 (left) and for N = 512 (right).
The spectrum in the temporally chaotic attractor (TCA) regime (ε = 0.199, dashed
line) is narrower than in the spatiotemporally chaotic attractor (STCA) regime (ε =
0.201, solid line). This indicates that at the onset of STC, when the spatial regularity
is destroyed, spectral energy cascades to neighboring modes due to nonlinear wave-
wave interactions, increasing the number of active modes. The energy spreading
remains essentially the same for N = 32 and N = 512.
The amount of spatial disorder can be quantified by means of the Fourier power
spectral entropy (REMPEL et al., 2007; POWELL; PERCIVAL, 1979; XI; GUNTON, 1995),
introduced in Subsection 2.2.2
SAk (t) = −N∑k=1
pk(t) ln(pk(t)), (4.10)
where pk(t) is the relative weight of a Fourier mode k at an instant t
pk(t) =|uk(t)|2∑Nk=1 |uk(t)|
2. (4.11)
96
0.196 0.198 0.2 0.2020.8
0.9
1
1.1
1.2
1.3
ε
TCS
STCS
Attractors
SA
k
FIGURE 4.3 - Time-averaged power spectral entropy as a function of the driver amplitudeε. The solid line represents the attractors (TCA and STCA), the dashedline denotes the spatiotemporally chaotic saddle (STCS), and the dottedline the temporally chaotic saddle (TCS).
SOURCE: Rempel et al. (2009)
Since u(x, t) in equation (4.4) is a real variable,
|u−k(t)| = |uk(t)|, (4.12)
only Fourier modes with k > 0 need to be considered. Note from Eq. (4.9) that the
mode u0(t) is decoupled from the other modes and is null for all t if u0(0) = 0. The
spectral entropy is maximum for a random system with uniform distribution, i.e.,
for all k, pk(t) = 1/N . In this case SAk (t) = lnN (BADII; POLITI, 1997). For N = 20,
the maximum entropy is ∼ 3.
Figure 4.3 shows the variation of the time-averaged power spectral entropy⟨SAk⟩
for
the attracting solutions (solid line) of Eq. (4.4) as a function of the control parameter
ε. The onset of spatiotemporal chaos can be clearly seen as a sudden increase in the
value of⟨SAk⟩
at ε ∼ 0.2. The other curves displayed in Fig. 4.3 are discussed in the
next section.
97
4.1.3 Transient and intermittent spatiotemporal dynamics
By neglecting the dissipation and forcing terms on the right-hand side of Eq. (4.4),
after multiplication by u and integration in the spatial domain, one can obtain an
equation to describe the temporal evolution of the“wave energy”, which is a constant
of motion for ν = ε = 0 (HE; SALAT, 1989; HE; CHIAN, 2005; BENJAMIN, 1972)
E(t) =1
4π
∫ 2π
0
[u2 − au2
x
]dx . (4.13)
The details of the derivation of this quantity can be found in Appendix D (Chapter
10) We use the time series of the wave energy to identify transient and intermittent
behaviors in the RLWE. Another quantity that has proven useful in this task is
the time series for the height of the main peak of the power spectrum h (REMPEL
et al., 2007). Figure 4.4 shows the time series for the wave energy E and the main
peak height h. In the TC regime (Fig. 4.4, left panels) the time series of E and h
display an initial behavior of high-level fluctuations before converging asymptotically
to a “laminar” state with lower variability, corresponding to the TCA of Fig. 4.1
(upper panel). In the STC regime (Fig. 4.4, right panels) the time series of E and
h display intermittent switchings between “bursty” (high variability) and “laminar”
(low variability) behaviors. It is important to distinguish the nature of this TC-STC
intermittency found in the RLWE from the spatiotemporal intermittency reported
by Chate and Manneville (1987), where patches of laminar and bursty behaviors
coexist in space. Here the bursty phases are clearly localized in time, but extend over
the whole space, similar to the spatiotemporal intermittency found in the damped
Kuramoto-Sivashinsky equation with periodic boundary conditions (REMPEL et al.,
2007) and experiments with liquid columns hanging below an overflowing circular
dish (BRUNET; LIMAT, 2004).
As mentioned before, chaotic transients are due to the presence of chaotic saddles in
the phase space. We adopt a Poincare map in which a point is plotted every time the
flow of Eqs. (4.9) crosses the plane Re(u1(t)) = 0 with d[Re(u1(t))]/dt > 0. Then,
we employ the sprinkler method (KANTZ; GRASSBERGER, 1985; HSU et al., 1988) to
find chaotic saddles. In Figure 4.5 the attracting and non-attracting chaotic sets are
represented as projections of the Poincare points on the [Re(u4),Re(u5)] plane. At
ε = 0.199, prior to the onset of STC, it is possible to find a chaotic saddle coexisting
with the temporally chaotic attractor (TCA). This chaotic saddle is responsible for
98
0
1
2
3
4
5
0 4000 80000
0.05
0.1
0.15
0 4000 8000
h
E
t t
STCATCASTCS
(ε = 0.199) (ε = 0.201)
FIGURE 4.4 - Time series of wave energy E (upper panels) and the maximum peak valueof power spectrum h (lower panels) for the regularized long wave equationin the temporally chaotic (TC, ε = 0.199, left panels) and spatiotemporallychaotic (STC, ε = 0.201, right panels) regimes. In the TC regime, thereis transient spatiotemporal chaos due to a spatiotemporally chaotic saddle(STCS).
SOURCE: Rempel et al. (2009)
99
transient spatiotemporal chaos, and is duly named spatiotemporally chaotic saddle
(STCS) (REMPEL; CHIAN, 2007; REMPEL et al., 2007). Figure 4.5 (upper panel) de-
picts the TCA (black) and the STCS (gray) for ε = 0.199. The latter surrounds
the region occupied by the TCA. After the transition to STC (ε = 0.201, Fig. 4.5
(middle panel)), the attractor expands abruptly to include the region previously oc-
cupied by the STCS. If one applies the sprinkler method in this regime, two chaotic
saddles embedded in the spatiotemporally chaotic attractor (STCA) can be found.
They are shown in Fig. 4.5, where the STCA (middle panel) is decomposed into a
spatiotemporally chaotic saddle (gray) and a temporally chaotic saddle (TCS, black)
(lower panel) which evolves from the destabilized TCA. In the sprinkler method, the
chaotic saddle is approximated by points from trajectories that follow long transients
before escaping from a predefined restraining region of the phase space. Given the
difficulty in defining a restraining region in a high-dimensional phase-space, we use
the different levels of wave energy displayed by the TC and STC regimes to identify
the “restraining regions”. To find the STC saddle, a large set of initial conditions is
iterated and those trajectories for which E(t) > 1.2 for 100 consecutive iterations of
the Poincare map (t ≈ 1000) are considered to be in the vicinity of the STC saddle.
For each of those trajectories, the first 40 and last 40 iterations are discarded and
only 20 points are plotted. For the TC saddle, the restraining region is defined as
above, but with E(t) < 1.2, instead. This threshold of E(t) is found after an in-
spection of the variability of E(t) in Fig. 4.4. The number of iterations discarded is
chosen after some trial-and-error (HSU et al., 1988).
The attractor decomposition mentioned in the previous paragraph suggests that
chaotic saddles dominate the spatiotemporal intermittent dynamics found after the
onset of STC. Their signatures can be seen in the time series shown in Fig. 4.4. The
STCS governs the dynamics of the initial transient in the left panel (TC regime),
and the bursty periods in the right panel (STC regime). Moreover, the similarity
between the TCA regime before the STC transition and the laminar phases after
the STC transition are due to the fact that, after the onset of STC, the temporally
chaotic attractor (TCA) loses its stability, becoming a temporally chaotic saddle
(TCS), which governs the laminar periods in the STCA regime. The TC-STC in-
termittency consists of random switchings between phases where the dynamics is
basically governed by TCS and STCS, respectively.
As the control parameter ε is increased beyond the transition point, the average
100
5u
Re(
)5
Re(
)u
5R
e(
)
u
TCA
ε = 0.201
TCS
STCA ε = 0.201
ε = 0.199STCS
STCS
4uRe( )
FIGURE 4.5 - Projections of attracting and non-attracting chaotic sets for the regularizedlong wave equation, for ε = 0.199 (upper panel), showing the temporallychaotic attractor (TCA, black) and the spatiotemporally chaotic saddle(STCS, gray); for ε = 0.201 (middle panel), showing the spatiotemporallychaotic attractor (STCA), which is decomposed in the lower panel intoa temporally chaotic saddle (TCS, black) and a spatiotemporally chaoticsaddle (STCS, gray).
SOURCE: Rempel et al. (2009)
101
log
10(τ
)
-4 -3.5 -3 -2.50
0.5
1
1.5
2
2.5
γ ∼ −1.05
10log ( )ε − ε
c
FIGURE 4.6 - Average duration of laminar intervals τ as a function of the departure fromthe critical value of the control parameter (εc = 0.2), in log-log scale, forthe regularized long-wave equation. The straight line shows a least-squaresfit with slope γ ∼ −1.05.
SOURCE: Rempel et al. (2009)
duration of the laminar periods τ decreases as a power law. In Fig. 4.6 we plot τ
as a function of the departure from the critical value of the control parameter (here
taken as εc = 0.2), in log-log scale. This result indicates that the STCS dominates
the spatiotemporal dynamics of the STCA when the value of the control parameter ε
is increased. This can also be seen in Fig. 4.3, where the values of the time-averaged
spectral entropy for STCS and TCS are represented by the dashed and dotted lines,
respectively. The power spectral entropy of STCA rapidly approaches the entropy of
STCS after the transition at εc ∼ 0.2. It is clear that the power spectral entropy of
STCS for ε < εc can be used to predict the dynamics in STCA for ε > εc as pointed
out in Rempel and Chian (2007).
4.2 Synchronization in the BBM equation
In this Section we investigate the amplitude and phase synchronization related with
multicale interactions after the onset of spatiotemporal chaos (on-off intermittency)
using the power-phase spectral entropies and the power-phase disorder parameters
introduced in Section 2.2. After the onset of spatiotemporal chaos (ε ∼ 0.2), a trajec-
tory on the spatiotemporally chaotic attractor that initially spends some time in the
102
vicinity of the spatiotemporally chaotic saddle (off-state in Figure 4.7) will escape
from it and approach the vicinity of the temporally chaotic saddle (on-state in Fig.
4.7). This process repeats itself back-and-forth, leading to an on-off spatiotemporal
intermittency seen in the time series of the “wave energy”E for ε = 0.20005 (upper
left-hand side panel of Fig. 4.7; see also left-hand side panels of Fig. 4.4). The middle
left-hand side panel show the Fourier power spectral entropy SAk and the bottom left-
hand side panel show the amplitude disorder parameter DAk . The upper right-hand
side panel shows the same time series of E, the middle right-hand side panel shows
the Fourier phase spectral entropy Sφk and the bottom right-hand side panel shows
the phase disorder parameter Dφk . The red lines denote the average taken within a
time window of 1256 time units to facilitate the visualization of the on-off states.
This Figure shows that the amplitude and phase dynamics are more synchronized
during the “on” stages (i.e. there is a lower degree of entropy and disorder) while
the “off” stages are characterized by a lower degree of synchronization (i.e. higher
degree of entropy and disorder). In particular, Sφk (t) shows that the transition from
the laminar (on) state to the bursty (off) state is much faster than the transition
from the bursty state to the laminar state.
In Section 4.1 it was demonstrated that the transient regimes in the time series of on-
off spatiotemporal intermittency at the onset of spatiotemporal chaos are related to
the chaotic saddles embedded in a chaotic attractor (REMPEL; CHIAN, 2007; REMPEL
et al., 2007; REMPEL et al., 2009). Fig 4.8 shows a three-dimensional projection of the
39-dimensional Poincare hyper-surface of section of chaotic attractors and chaotic
saddles found before and after transition. Figure 4.8(a) shows the STCS detected
by using the sprinkler method for ε = 0.199, before the onset of spatiotemporal
chaos. This chaotic saddle governs the dynamics of transients before converging to
the temporally chaotic attractor (TCA) which is shown in Fig. 4.8(b). Note that
the TCA occupies a smaller volume of phase space, which resembles a surface. In
contrast, the STCS occupies a larger, sphere-like volume of phase space. At ε ∼ 0.2,
a crisis occurs due to a chaotic attractor-chaotic saddle collision involving the TCA
and the STCS. At crisis, the TCA loses its stability and is converted to a temporally
chaotic saddle (TCS), as seen in Fig. 4.8(c) for ε = 0.20005; the STCS is robust and
persists after crisis, also shown in Fig. 4.8(c). At the onset of spatiotemporal chaos,
the TCS becomes coupled to the STCS through the coupling unstable periodic or-
bits to form a STCA, as seen in Fig. 4.8(d), which occupies almost the same region
of phase space as the pre-crisis STCS and TCA. The TCS and the STCS constitute
103
On off spatiotemporal intermittency
0.0
3.0
6.0
0.0
1.0
2.0
0.0
0.5
1.0
0.0
3.0
6.0
0.0
4.0
8.0
0.0
1.0
2.0
AD
k
Dφ k
= 0.20005)ε(STCA,
Skφ
SA k
E
Phase dynamicsoffoff
on
offoff
on
off
on on
off
Amplitude dynamics
E
0 70000
t0 7000035000
t35000
FIGURE 4.7 - Amplitude and phase dynamics of the spatiotemporally chaotic attractor(STCA) after the crisis-like transition (ε = 0.20005). Left panels show fromtop to bottom: wave energy E, power spectral entropy SAk and averagedwave number DA
k . Right-hand panels show from top to bottom: wave energyE, phase spectral entropy Sφk and disorder parameter Dφ
k . The red linesdenote the average over 1256 time units (∼ 816 driver periods).
the “skeleton” of the STCA. Since this value of ε is close to the transition point,
the STCA keeps a “memory” of the former TCA, which can be seen as a region
with higher concentration of points within the STCA. An analysis of Figs. 4.8(b)
and 4.8(d) suggests the occurrence of a blowout bifurcation (FUJISAKA; YAMADA,
1983) at the transition from the TCA to the STCA. Prior to the blowout bifur-
cation, trajectories of TCA are confined to a synchronization manifold represented
by the sheet-like structure of Fig. 4.8(b). At the onset of the blowout bifurcation,
trajectories of STCA lose their transverse stability owing to the unstable periodic
orbits in STCS and traverse across the sheet-like region, permeating the sphere-like
structure. The asymmetry between on-off and off-on transitions in Sφk (t) in Fig. 4.7
is caused by the asymmetry of the trajectory dynamics traversing from the vicinity
of TCS to the vicinity of STCS and vice-versa.
104
3R
eu
3R
eu
3R
eu
3
Re
u
(ε = 0.199)(ε = 0.199)TCA(a) (b)STCS
Poincare plots of chaotic saddles & chaotic attractors´
(ε = 0.20005)STCS + TCS(c) STCA(d)
0.1
0.2 0.1
0.1 0.2
0.1
0.1
0.2 0.1
2Reu 2
Imu
0.1
0.0
0.1
0.1
0.2 0.1
2Reu
2
Imu
0.1
2Reu
2
Imu
2Reu
2
Imu
0.1
0.0
0.1
0.10.0
0.10.0
0.0
0.1
0.1
0.10.0
0.1
0.10.0
0.00.1
0.0
0.10.0
0.10.0
0.0
0.1
(ε = 0.20005)
FIGURE 4.8 - (a) Three-dimensional projection of the Poincare hyper-surface of section ofthe spatiotemporal chaotic saddle (STCS) previous to the crisis-like tran-sition. (b) The temporally chaotic atrractor (TCA) previous to the tran-sition. (c) The temporally chaotic saddle (TCS, blue) and STCS (yellow)after transition. (d) The spatiotemporally chaotic attractor (STCA) aftertransition.
105
0.0
2.0
4.0
0.0
1.0
2.0
0.0
4.0
8.0
0.0
1.0
2.0
0.0
0.5
1.0
ADk
Dφ k
SA k
Skφ
E
0 12000 024000 12000 24000
t t
ε( = 0.20005)
Amplitude phase dynamics of chaotic saddles
TCS STCS
FIGURE 4.9 - Amplitude and phase dynamics of the temporally chaotic saddle (TCS, leftpanels) and the spatiotemporally chaotic saddle (STCS, right panels) afterthe crisis-like transition (ε = 0.20005). From top to bottom: wave energyE, spectral entropy SAk , averaged wave number DA
k , phase entropy Sφk anddisorder parameter Dφ
k . The red lines denote the average over 1256 timeunits (∼ 816 driver periods).
106
The degree of amplitude-phase synchronization in the chaotic saddles (TCS and
STCS) embedded in the STCA at the onset of spatiotemporal chaos can be deter-
mined by calculating the Fourier power and phase entropies and the amplitude-phase
disorder parameters. We used the stagger-and-step procedure (SWEET et al., 2001)
for the computation of long trajectories which approximate chaotic saddles in a
high-dimensional phase space. This method allows us to obtain an arbitrarily long
pseudo-trajectory of a chaotic saddle. Figure 4.9 shows the amplitude and phase
dynamics of the chaotic saddles found in ε = 0.20005. The left-hand side panels
show, from top to bottom, the time series of “wave energy” E , the Fourier power
spectral entropy SAk , the amplitude disorder parameter DAk , the Fourier phase spec-
tral entropy Sφk and the phase disorder parameter Dφk of TCS. The right-hand side
panels show, from top to bottom, the time series of “wave energy” E, the Fourier
power spectral entropy SAk , the amplitude disorder parameter DAk , the Fourier phase
spectral entropy Sφk and the phase disorder parameter Dφk of STCS. The red line
denotes the averaging over 1256 time units. A comparison of Figs. 4.7 and 4.9 show
that the average values of SAk ∼ 0.93, DAk ∼ 3.1, Sφk ∼ 1.01 and Dφ
k ∼ 0.31 of the
TCS in Fig. 4.9 are close to the corresponding values of the on-state of Fig. 4.7;
moreover, the average values of SAk ∼ 1.48, DAk ∼ 4.65, Sφk ∼ 1.67 and Dφ
k ∼ 0.75
of the STCS in Fig. 4.9 are close to the corresponding values of the off-state of Fig.
4.7. This confirms that the on-off states of the spatiotemporal intermittency at the
onset of permanent chaos correspond to the system trajectory traversing the vicinity
of TCS and STCS, respectively.
The energy distribution among Fourier modes in the k wavenumber domain can be
examined by constructing the time-averaged power spectrum⟨|uk|2
⟩. In addition,
we computed the time-averaged phase-difference spectrum 〈|δφk|〉 as a qualitative
measurement of synchronization among Fourier modes. Smaller values of 〈|δφk|〉 cor-
respond to a higher degree of synchronization between Fourier modes k and k + 1.
Fig. 4.10 shows both Fourier power and phase-difference spectra for ε = 0.21, corre-
sponding to the STC regime, for the STCA (red), STCS (yellow) and TCS (blue).
The power spectrum of both STCA and STCS are broader than the TCS, indicating
that the energy is spread among Fourier modes, while the TCS has its energy con-
centrated in a narrow band of wavenumbers. The phase-difference spectra of STCA
and STCS display a tendency to increase with k, suggesting that synchronization
is lower at high wavenumbers. The phase-difference spectra of TCS is, in general,
lower than the STCA and STCS, which is consistent with the fact that the TCS
107
5 10 15 20
10-3
10-2
10-1
STCASTCSTCS
5 10 150
1×10-3
2×10-3
3×10-3
4×10-3
5×10-3
k
2|u | δφ | |
k
Power spectrum Phase spectrum
k k
FIGURE 4.10 - Left panel: time-averaged power spectra⟨|uk|2
⟩for ε = 0.21, of the spa-
tiotemporally chaotic attractor (STCA, red thin line), the spatiotempo-rally chaotic saddle (STCS, yellow thick line) and the temporally chaoticsaddle (TCS, blue thick line). Right panel: time-averaged phase-differencespectra 〈|δφk|〉 of STCA, STCS and TCS.
represents synchronized (“on”) states.
Next, we compute the Lyapunov spectrum of chaotic sets (YAMADA; OKHITANI,
1988; YAMADA; OHKITANI, 1998) before and after the onset of spatiotemporal chaos.
We use the method of Gram-Schmidt orthogonalization (SHIMADA; NAGASHIMA,
1979) and order the Lyapunov exponents as λj > λj+1. The spectrum of rescaled
Lyapunov exponents λj/H of chaotic attractors and chaotic saddles before and after
transition are shown in the top panel of Fig. 4.11, where H denotes the Kolmogorov-
Sinai entropy (Equation (2.43)). Before transition (ε = 0.199), the TCA (green)
has only one positive Lyapunov exponent (λ1 ∼ 0.052418), whereas the pre-crisis
STCS (orange) has 14 positive Lyapunov exponents. After transition (ε = 0.21), the
Lyapunov spectra of both the STCA (red) and the post-crisis STCS (yellow) display
14 positive Lyapunov exponents, whereas the TCS (blue) only has one positive
exponent (λ1 ∼ 0.061873). The similarity between λ1 of TCA before transition and
TCS after transition is due to the fact that, after the onset of STC (ε ∼ 0.2), the
TCA loses its stability becoming a TCS. At ε = 0.21, the Lyapunov spectrum of
STCA has numerical values almost identical to the spectrum of STCS, except for
j ≥ 12. The difference between the Lyapunov spectra of STCA and STCS for j ≥ 12
can be attributed to the fact that both TCS and STCS contribute to the dynamics
of STCA, hence the Lyapunov spectra of STCA is a nonlinear combination of both
TCS and STCS.
108
0 5 10 1510-4
10-2
100
λ j/H
TCA (ε = 0.199)STCS (ε = 0.199)
(a)
(b)
STCA (ε = 0.21)STCS (ε = 0.21)TCS (ε = 0.21)
0 10 20 30 401
2
3
4
5
⟨Sj⟩
j0 = 2
Aφ
STCA (ε = 0.21)STCS (ε = 0.21)TCS (ε = 0.21)
0 10 20 30 40j
1
2
3
⟨Sj⟩
STCA (ε = 0.21)STCS (ε = 0.21)TCS (ε = 0.21)
(c)
FIGURE 4.11 - Top panel: spectrum of positive rescaled Lyapunov exponents λj/H as afunction of Lyapunov index j of the temporally chaotic attractor (TCA)and the spatiotemporally chaotic saddle (STCS) for ε = 0.199, and spa-tiotemporal chaotic attractor (STCA), the spatiotemporally chaotic sad-dle (STCS) and the temporally chaotic saddle (TCS) for ε = 0.21. Hdenotes the Kolmogorov-Sinai entropy (H =
∑qj=1 λj , |λq > 0, λq+1 ≤ 0).
Middle panel: power spectral entropy as a function of Lyapunov index j ofSTCA, STCS and TCS for ε = 0.21. Bottom panel: phase spectral entropyas a function of Lyapunov index j of STCA, STCS and TCS for ε = 0.21.Arrows indicate the Lyapunov index in which λj=j0 ∼ 0 for the TCS.
109
TABLE 4.1 - Kolmogorov-Sinai entropy H and Kaplan-Yorke dimension D of the on-offspatiotemporal intermittency at ε = 0.21 for the embedded spatiotemporallychaotic saddle (STCS) and temporally chaotic saddle (TCS).
STCS TCS
H 0.38 0.06
D 36.15 22.23
The Lyapunov vectors are embedded in a 40-dimensional phase space, each di-
mension corresponds to either the real or imaginary part of the complex Fourier
mode uk, k = 1, ..., 20. Therefore, it is possible to obtain the power spectrum
and phase spectrum of each Lyapunov vector as a function of j and k (see Sub-
section 2.2.3.1). The power and phase spectral entropy as a function of j can be
calculated by using the information of amplitudes and phases. The middle panel
of Fig. 4.11 shows the time-averaged Lyapunov power spectral entropy⟨SAj⟩
=⟨−∑N
k=1 p(δujk(t)) ln[p(δujk(t))]
⟩, and the bottom panel displays the time-averaged
Lyapunov phase spectral entropy⟨Sφj
⟩=⟨−∑N
k=1 P (δφjk(t)) ln[P (δφjk(t))]⟩
as a
function of Lyapunov index j for STCA, STCS and TCS at ε = 0.21. Arrows indi-
cate the Lyapunov index corresponding to the zero Lyapunov exponent for the TCS
(j0 = 2). It is evident that the power and phase entropies of TCS are lower than
STCA and STCS for all j. At smaller j,⟨SAj⟩
and⟨Sφj
⟩decrease, which indicates
that, in average, the power spectra of the first Lyapunov vectors are concentrated in
a narrow band of wavenumbers, and their phase spectra are more synchronized. Note
that, in the numerical simulations, energy is injected into the k = 1 Fourier mode.
The above result of⟨SAj (t)
⟩and
⟨Sφj (t)
⟩is in agreement with the Kolmogorov-Sinai
entropy H and the Kaplan-Yorke dimension D (Eq. (2.44)) computed from the Lya-
punov spectrum (YAMADA; OKHITANI, 1988) for both STCS and TCS, shown in
Table 4.1. For STCA, the numerical values of H and D are very close to the respec-
tive values of STCS.
The time-averaged Fourier-Lyapunov power spectra⟨|δujk|2
⟩of the STCA, the STCS
and the TCS at ε = 0.21 are shown in the left-hand side of Fig. 4.12. Colour levels
110
are indicated in the scales at the bottom of the Figures. For STCA and STCS, j = 15
corresponds to the zero Lyapunov exponent. The Fourier-Lyapunov power spectra of
STCA and STCS display a maximum near the origin of coordinates (j = 0, k = 0).
This means that the first Lyapunov vectors, which are associated with the most
unstable directions, have their energy concentrated at smaller wavenumbers. As a
consequence, there is a strong correspondence between the first Lyapunov vectors
and small wavenumbers, i.e. the Fourier and Lyapunov bases are “frozen” to each
other near the origin (OHKITANI; YAMADA, 1989). Both STCA and STCS have
their energy confined in a narrow inverted-V region in the Fourier-Lyapunov space,
distinct from the Gledzer-Ohkitani-Yamada (GOY) shell model of fully-developed
turbulence where the energy is confined in a narrow V region (YAMADA; OHKITANI,
1998). The power-Lyapunov spectrum of TCS (Fig. 4.12, bottom panel) also shows
a peak near the origin, but the energy spreading is not seen, because the TCS has
only one positive Lyapunov exponent. For all j, the TCS display a higher energy
localization than STCA and STCS, which is in agreement with a low degree of
power spectral entropy as shown in the middle panel of Figure 4.11. Note that
for the TCS, the large energy peaks observed for j > j0 corresponds to negative
Lyapunov exponents, which are related to the dissipation of energy. The smaller
left-hand side panels show the corresponding time-averaged Fourier power spectra
〈|uk|2〉 as a function of wavenumber k, which confirm that the energy bandwidth of
the TCS is narrower than the bandwidth of the STCS and STCA.
The difference between the Fourier-Lyapunov power spectrum of STCA and STCS
shown in Figure 4.12, and the Fourier-Lyapunov power spectrum of the GOY shell
model (YAMADA; OHKITANI, 1998) arises from the way energy injection and dissipa-
tion are modelled. In Yamada and Ohkitani (1998) energy is injected in the fourth
shell, which is analogous to the fourth Fourier mode, whereas the BBM equation
energy is injected into the first Fourier mode. The energy dissipation mechanisms
are also different; Figure 4.13 shows the dissipation term of the BBM equation and
the shell model as a function of wavenumber k.
In addition to the Fourier-Lyapunov power spectrum, the time-averaged phase-
differences spectra⟨δφjk⟩
of STCA, STCS and TCS are displayed in the right-hand
side panels of Fig. 4.12. The Fourier-Lyapunov phase spectra of STCA and STCS
display large peaks of⟨δφjk⟩
corresponding to nonsynchronized modes at all ks and
all js. In contrast, the phase differences spectrum of TCS are small across all ks at
111
0
10
20
j
k
2| u |δ
0 20 30 4010
0 20 30 4010
0 20 30 4010
0.01 0.1
0.01 0.1
0.01 0.1
|u |k
2
(a)
(b)
(c)
(d)
(e)
(f)
δφ jk
0 20 30 4010
0 20 30 4010
0 20 30 4010
0.0 0.004
0.0 0.004
0.0 0.004
kδφ
Fourier Lyapunov power & phase spectra
STCA
j
k
TCS
0
10
20
k
STCS
0
10
20
j
k
0.10.0 0.2
j
STCA
0
10
20
k
j
TCS
0
10
20
k
STCS
0
10
20
k
j
0.10.00.1
j
Amplitude dynamics Phase dynamics
FIGURE 4.12 - Left-hand side panels: Time average of power-Lyapunov spectrum⟨|δujk|
2⟩
of the spatiotemporally chaotic attractor (STCA, top panel),the spatiotemporally chaotic saddle (STCS, middle panel) and the tem-porally chaotic saddle (TCS, bottom panel) at ε = 0.21. Right-hand sidepanels: time average of phase-Lyapunov spectrum of STCA (top panel),STCS (middle panel) and TCS (bottom panel). Fourier wavenumber isindicated by k, and j represent Lyapunov indices. Lyapunov vectors areordered from higher to lower values of their respective Lyapunov expo-nents. The side panels show the time-averaged Fourier power spectra andthe time averaged Fourier phase spectra, respectively.
112
100
101
k
10-4
10-3
10-2
10-1
ν /
(1 −
ak2)
BBM Equation
10-1
100
101
k
10-9
10-6
νk2
GOY Shell Model
FIGURE 4.13 - Dissipation profiles of the Benjamin-Bona-Mahony equation (left sidepanel) and the Shell model of turbulence (right side panel) as a functionof wavenumber k.
low js. Note that the Fourier modes of the unstable Lyapunov vector of TCS (j = 1)
are synchronized, whereas the STCA and STCS display large phase differences for
j < 15, corresponding to unstable Lyapunov vectors. The smaller right-hand side
panels show the plots of the corresponding time-averaged Fourier phase spectra.
These results are in agreement with the bottom panel of Fig. 4.11, in which lower
degree of the phase spectral entropy (i.e. higher degree of synchronization) was ob-
served for the TCS.
113
5 CONCLUSION
In Subsection 3.1 we characterized intermittency and phase synchronization in inter-
mittent magnetic field turbulence. The results presented in subsection 3.1.1 provide
the first observational proof of phase coherence in the ambient solar wind turbu-
lence, based on the magnetic field data of ACE at L1. Figure 3.12 indicates that the
level of intermittency and phase synchronization detected by both Cluster and ACE
are very similar except for scales around 10 s, where ACE observed a higher level of
intermittency and phase synchronization than Cluster. The peak regions of kurtosis
and phase coherence index in Figure 3.12 corresponds to the spectral break regions
in Figure 3.6 where the magnetic field turbulence is dominated by nonlinear wave-
wave and wave-particle interactions. Since Cluster is located in the shocked solar
wind, the reflected ions from the Earth’s bow shock can enhance the dissipation of
nonlinear Alfven waves via ion-cyclotron damping and other kinetic effects (HOWES
et al., 2008), leading to a decrease of phase synchronization.
Our study based on Cluster and ACE observations demonstrate that the intermit-
tency in the magnetic field turbulence, in the shocked solar wind upstream of Earth’s
bow shock and in the unshocked ambient solar wind at L1, is the result of synchro-
nization intrinsic in nonlinear multiscale interactions. Numerical simulations of non-
linear plasma waves have confirmed that intermittent events are localized regions of
plasmas or fluids governed by bursts of energy spikes (coherent structures) where
phase synchronization is operating (HE; CHIAN, 2003; HE; CHIAN, 2005). Since large-
amplitude coherent structures of small scales have typical lifetimes longer than that
of small-amplitude incoherent (stochastic) fluctuations, the dynamics of an intermit-
tent turbulence, ubiquitous in the heliophysical environment (CHIAN et al., 2006), is
dominated by coherent structures resulting from amplitude-phase synchronization.
In subsection 3.1.2 we extended our study presented in subsection 3.1.1 to include an
ICME event observed in January 2005. We analyzed magnetic field measurements
from SOHO MDI solar images, ACE and Cluster data in the solar wind, and ground
magnetometers in Brazil. Finite degree of non-Gaussianity and phase synchroniza-
tion were observed in all datasets. We showed that the interplanetary magnetic field
turbulence downstream of the ICME shock was closely correlated with the Earth’s
geomagnetic turbulence detected on the ground. In a future work we will extend our
analysis of solar images using high-resolution Hinode data, which will allow us to in-
vestigate non-Gaussianity and synchronization at smaller spatial scales than SOHO
115
magnetograms. Some preliminar results are presented in Appendix A (Chapter 7).
The turbulent exchange of mass and momentum from and within canopies is domi-
nated by coherent structures (FINNIGAN, 2000). Wesson et al. (2003) applied three
nonlinear time series techniques (Shannon entropy, wavelet thresholding, and mutual
information content) to contrast the level of organization in vertical wind velocity
in the canopy sublayer and the atmospheric surface layer. In Section 3.2 we have
demonstrated that both kurtosis and phase coherence index techniques are capable
of characterizing the degree of departure from Gaussianity, due to phase coherence,
of atmospheric intermittent turbulence. The nonlinear techniques discussed can be
applied to investigate the role played by coherent structures in experimental (GAO
et al., 1989; BARTHLOTT et al., 2007), theoretical (RAUPACH et al., 1996; HARMAN;
FINNIGAN, 2007), and large-eddy simulation (SU et al., 1998; QIU et al., 2008) studies
of atmospheric turbulence in forest canopy, as well as in orchard canopy (WANG et
al., 1992; STOUGHTON et al., 2002), rice canopy (GAO et al., 2003), corn canopy (YUE
et al., 2007; ZHU et al., 2007), cotton and grape canopies (MITIC et al., 1999), coral
canopy (REIDENBACH et al., 2007), and urban canopy (FEIGENWINTER; VOGT, 2005;
SALMOND et al., 2005).
The theoretical results from numerical simulations of the Benjamin-Bona-Mahony
(BBM) equation (Section 4.1) are consistent with those reported by Rempel et al.
(2007), where the damped Kuramoto-Sivashinsky equation with periodic boundary
conditions was studied. We suggest that the mechanism for the onset of TC-STC
intermittency via the coupling of chaotic saddles can be readily found in other fluid
systems in transition from laminar to weakly turbulent flows, provided the following
conditions are met:
a) There is a discontinuous transition from temporal chaos to spatiotemporal
chaos, due to a crisis-like phenomenon.
b) Before the transition, the system displays transient spatiotemporal chaos.
Examples of fluids with chaotic transients and a crisis transition to spatiotemporal
chaos or turbulence may include the onset of bursting behavior in a driven, two-
dimensional viscous flow subject to no-slip boundaries (MOLENAAR et al., 2007) and
the onset of turbulence in pipe flows. In the latter case, a boundary crisis seems to
be responsible for converting a transient turbulent state into a turbulent attractor at
116
Reynolds ∼ 1800 (WILLIS; KERSWELL, 2007; PEIXINHO; MULLIN, 2006). Although
recent experiments seem to indicate that turbulence in pipe flows is indeed a tran-
sient phenomenon (HOF et al., 2006; HOF et al., 2008), thus contradicting the previous
results. Whether it is transient for all Reynolds numbers or there is a crisis bifurca-
tion to an attractor remains an open question (ECKHARDT, 2008).
In Section 4.2 we used the mathematical tools described in Section 2.2 to measure
the degree of amplitude and phase synchronization in numerical simulations of the
BBM equation after the onset of spatiotemporal chaos. We computed the power-
phase spectral entropies and the power-phase disorder parameters using Fourier and
Lyapunov representations to show that the laminar and bursty periods in the on-
off spatiotemporal intermittency correspond, respectively, to the temporally chaotic
saddle with higher degree of amplitude-phase synchronization and the spatiotem-
porally chaotic saddle with lower degree of amplitude-phase synchronization across
spatial scales.
As a conclusion, the observational results from Chapter 3 indicate that kurtosis
and the phase coherence index display similar variations across scales. The theoret-
ical results from Chapter 4 demonstrates that higher/lower degree of power spec-
tral entropy correspond to higher/lower degree of phase spectral entropies, which
is consistent with our observational results. This duality of amplitude and phase
synchronization may be the origin of intermittency in fully-developed turbulence in
the solar-terrestrial environment.
117
REFERENCES
ABRAMENKO, V.; YURCHYSHYN, V.; WANG, H. Intermittency in the
photosphere and corona above an active region. The Astrophysical Journal,
v. 681, p. 1669–1676, 2008. 135
ABRAMENKO, V. I.; YURCHIYSHIYN, V. B.; WANG, H.; SPIROCK, T. J.;
GOODE, P. R. Scaling behavior of structure functions of the longitudinal magnetic
field in active regions on the sun. The Astrophysical Journal, v. 577, p.
487–495, 2002. 74, 135
ALEXANDROVA, O. Solar wind vs magnetosheath turbulence and Alfven
vortices. Nonlinear Processes in Geophysics, v. 15, p. 95–108, 2008. 63
ALEXANDROVA, O.; CARBONE, V.; VELTRI, P.; SORRISO-VALVO, L. Solar
wind Cluster observations: turbulent spectrum and role of Hall effect. Planetary
and Space Science, v. 55, p. 2224–2227, 2007. 53, 54
. Small-scale energy cascade of the solar wind turbulence. The
Astrophysical Journal, v. 674, p. 1153–1157, 2008. 54, 60, 63, 72, 88
ALLIGOOD, K. T.; SAUER, T. D.; YORKE, J. A. Chaos: an introduction to
dynamical systems. New York: Springer-Verlag, 1996. 35
BADII, R.; POLITI, A. Complexity: hierarchical structures and scaling in
physics. Cambridge: Cambridge University Press, 1997. 97
BALE, S. D.; BALIKHIN, M. A.; HORBURY, T. S.; KRASNOSELSKIKH, V. V.;
KUCHAREK, H.; MOBIUS, E.; WALKER, S. N.; BALOGH, A.; BURGESS, D.;
LEMBEGE, B.; LUCEK, E. A.; SCHOLER, M.; SCHWARTZ, S. J.; THOMSEN,
M. F. Quasi-perpendicular shock structure and processes. Space Science
Reviews, v. 118, p. 161–203, 2005a. 55
BALE, S. D.; KELLOGG, P. J.; MOZER, F. S.; HORBURY, T. S.; REME, H.
Measurement of the electric fluctuation spectrum of magnetohydrodynamic
turbulence. Physical Review Letters, v. 94, n. 215002, 2005b. 53
BALOGH, A.; CARR, C. M.; NA, M. H. A.; DUNLOP, M. W.; BEEK, T. J.;
BROWN, P.; FORNACON, K.-H.; GEORGESCU, E.; GLASSMEIER, K.-H.;
HARRIS, J.; MUSMANN, G.; ODDY, T.; SCHWINGENSCHUH, K. The Cluster
119
magnetic field investigation: overview of in-flight performance and initial results.
Annales Geophysicae, v. 19, p. 1207–1217, 2001. 57
BARTHLOTT, C.; DROBINSKI, P.; FESQUET, C.; DUBOS, T.; PIETRAS, C.
Long-term study of coherent structures in the atmospheric surface layer.
Boundary-Layer Meteorology, v. 125, p. 1–24, 2007. 116
BEHLKE, R.; ANDRE, M.; BALE, S. D.; PICKETT, J. S.; CATTELL, C. A.;
LUCEK, E. A.; BALOGH, A. Solitary structures associated with short
large-amplitude magnetic structures (SLAMS) upstream of the Earth’s
quasi-parallel bow shock. Geophysical Research Letters, v. 31, p. L16805,
2004. 56, 87
BENJAMIN, T. B. The stability of solitary waves. Proceedings of the Royal
Society of London Series A, v. 328, p. 153–183, 1972. 98, 149, 150, 151
BENJAMIN, T. B.; BONA, J. L.; MAHONY, J. J. Model equation for long waves
in nonlinear dispersive systems. Philosophical Transactions of the Royal
Society of London Series A, v. 272, p. 47–78, 1972. 91, 92, 93, 149, 151
BENZI, R.; CILIBERTO, S.; TRIPICCIONE, R.; BAUDET, C.; MASSAIOLI, F.;
SUCCI, S. Extended self-similarity in turbulent flows. Physical Review E, v. 48,
p. R29–R32, 1993. 69
BISKAMP, D. Magnetohydrodynamic turbulence. Cambridge: Cambridge
University Press, 2003. 26, 143
BISKAMP, D.; SCHWARZ, E.; ZEILER, A.; CELANI, A.; DRAKE, J. F.
Electron magnetohydrodynamic turbulence. Physics of Plasmas, v. 6, n. 3, p.
751–758, 1999. 61
BOHR, T.; JENSEN, M. H.; PALADIN, G.; VULPIANI, A. Dynamical systems
approach to turbulence. Cambridge: Cambridge University Press, 1998. 31
BOLZAN, M. J.; RAMOS, F. M.; SA, L. D. A.; NETO, C. R.; ROSA, R. R.
Analysis of fine-scale canopy turbulence within and above an Amazon forest using
Tsallis’ generalized thermostatistics. Journal of Geophysical Research -
Atmospheres, v. 107, n. D20, p. 8063, 2002. 81
120
BOROVSKY, J. E. Flux tube texture of the solar wind: strands of the magnetic
carpet at 1 AU? Journal of Geophysical Research - Space Physics, v. 113,
p. A08110, 2008. 87, 88
BRAUN, R.; FEUDEL, F. Supertransient chaos in the two-dimensional complex
Ginzburg-Landau equation. Physical Review E, v. 53, p. 6562–6565, 1996. 36
BRUNET, P.; LIMAT, L. Defects and spatiotemporal disorder in a pattern of
falling liquid columns. Physical Review E, v. 70, p. 046207, 2004. 98
BRUNO, R.; CARBONE, V. The solar wind as a turbulence laboratory. Living
Reviews in Solar Physics, v. 2, n. 4, 2005. 51, 60
BRUNO, R.; CARBONE, V.; BAVASSANO, B.; SORRISO-VALVO, L.
Observations of magnetohydrodynamic turbulence in the 3D heliosphere.
Advances in Space Research, v. 35, p. 939–950, 2005. 60, 87, 88
BRUNO, R.; CARBONE, V.; SORRISO-VALVO, L.; BAVASSANO, B. Radial
evolution of solar wind intermittency in the inner heliosphere. Journal of
Geophysical Research - Space Physics, v. 108, n. A3, p. 1130, 2003. 57, 86,
87, 88
BRUNO, R.; CARBONE, V.; VELTRI, P.; PIETROPAOLO, E.; BAVASSANO, B.
Identifying intermittency events in the solar wind. Planetary and Space
Science, v. 49, p. 1201–1210, 2001. 87, 88
BURGESS, D.; LUCEK, E. A.; SCHOLER, M.; BALE, S. D.; BALIKHIN, M. A.;
BALOGH, A.; HORBURY, T. S.; KRASNOSELSKIKH, V. V.; KUCHAREK, H.;
LEMBEGE, B.; MOBIUS, E.; SCHWARTZ, S. J.; THOMSEN, M. F.; WALKER,
S. N. Quasi-parallel shock structure and processes. Space Science Reviews,
v. 118, p. 205–222, 2005. 56
BURLAGA, L. F.; VINAS, A. F. Multi-scale probability distributions of solar
wind speed fluctuations at 1 AU described by a generalized Tsallis distribution.
Geophysical Research Letters, v. 31, p. L16807, 2004. 54
CAKMUR, R. V.; EGOLF, D. A.; PLAPP, B. B.; BODENSCHATZ, E. Bistability
and competition of spatiotemporal chaotic and fixed point attractors in
Rayleigh-Benard convection. Physical Review Letters, v. 79, p. 1853–1856,
1997. 39
121
CAMPANHARO, A. S. L. O.; RAMOS, F. M.; MACAU, E. E. N.; ROSA, R. R.;
BOLZAN, M. J. A.; SA, L. D. A. Searching chaos and coherent structures in the
atmospheric turbulence above the Amazon forest. Philosophical Transactions
of the Royal Society of London - Mathematical and Physical Sciences,
v. 366, p. 579–589, 2008. 89
CHANG, T.; TAM, S.; WU, C.-C. Complexity induced anisotropic bimodal
intermittent turbulence in space plasmas. Physics of Plasmas, v. 11, n. 4, p.
1287–1299, 2004. 88
CHAPMAN, S. C.; HNAT, B. Quantifying scaling in the velocity field of the
anisotropic turbulent solar wind. Geophysical Research Letters, v. 34, p.
L17103, 2007. 54
CHATE, H.; MANNEVILLE, P. Transition to turbulence via spatiotemporal
intermittency. Physical Review Letters, v. 58, n. 2, p. 112–115, 1987. 98
CHIAN, A. C.-L.; KAMIDE, Y.; REMPEL, E. L.; SANTANA, W. M. On the
chaotic nature of solar-terrestrial environment: interplanetary Alfven
intermittency. Journal of Geophysical Research - Space Physics, v. 111, n.
A07S03, 2006. 89, 115
CHIAN, A. C.-L.; MIRANDA, R. A. Cluster and ACE observations of phase
synchronization in intermittent magnetic field turbulence: a comparative study of
shocked and unshocked solar wind. Annales Geophysicae, v. 27, p. 1789–1801,
2009. 56, 57, 59, 61, 62, 65, 67, 68, 70, 71, 74, 75, 135
CHIAN, A. C.-L.; MIRANDA, R. A.; KOGA, D.; BOLZAN, M. J. A.; RAMOS,
F. M.; REMPEL, E. L. Analysis of phase coherence in fully developed atmospheric
turbulence: Amazon forest canopy. Nonlinear Processes in Geophysics, v. 15,
p. 567–573, 2008. 33, 51, 69, 70, 80, 81, 82, 83, 84, 85, 86
CHIANG, L.-Y.; COLES, P. Phase information and the evolution of cosmological
density perturbations. Monthly Notices of the Royal Astronomical Society,
v. 311, p. 809–824, 2000. 40
DAUBECHIES, I. Ten lectures on wavelets. Philadelphia: Society for Industrial
and Applied Mathematics (SIAM), 1994. 64
DAVIDSON, P. A. Turbulence: an introduction for scientists and
engineers. New York: Oxford University Press, 2004. 25, 31, 65, 69, 85, 142
122
DE WIT, T. D. Can high-order moments be meaningfully estimated from
experimental turbulence measurements? Physical Review E, v. 70, n. 055302,
2004. 32, 64, 66, 69
DE WIT, T. D.; KRASNOSELSKIKH, V. V. Non-Gaussian statistics in space
plasma turbulence: fractal properties and pitfalls. Nonlinear Processes in
Geophysics, v. 3, p. 262–273, 1996. 66
DODD, R. K.; EILBECK, J. C.; GOBBON, J. D.; MORRIS, H. C. Solitons and
nonlinear wave equations. London: Academic Press, 1982. 91
EASTWOOD, J. P.; BALOGH, A.; LUCEK, E. A.; MAZELLE, C.;
DANDOURAS, I. On the existence of Alfven waves in the terrestrial foreshock.
Annales Geophysicae, v. 21, p. 1457–1465, 2003. 56, 58, 72
ECKHARDT, B. Turbulence transition in pipe flow: some open questions.
Nonlinearity, v. 21, p. T1–T11, 2008. 117
FEIGENWINTER, C.; VOGT, R. Detection and analysis of coherent structures in
urban turbulence. Theoretical and Applied Climatology, v. 81, p. 219–230,
2005. 116
FINNIGAN, J. Turbulence in plant canopies. Annual Review of Fluid
Mechanics, v. 32, p. 519–571, 2000. 116
FOULLON, C.; OWEN, C. J.; DASSO, S.; GREEN, L. M.; DANDOURAS, I.;
ELLIOTT, H. A.; FAZAKERLEY, A. N.; BOGDANOVA, Y. V.; CROOKER,
N. U. Multi-spacecraft study of the 21 january 2005 ICME: evidence of current
sheet substructure near the periphery of a strongly expanding, fast magnetic cloud.
Solar Physics, v. 244, p. 139–165, 2007. 75
FRISCH, U. Turbulence - the legacy of A. N. Kolmogorov. Cambridge:
Cambridge University Press, 1995. 26, 31, 33, 66, 69, 85, 142
FUJISAKA, H.; YAMADA, T. Stability theory of synchronized motion in coupled
oscillator systems. Progress of Theoretical Physics, v. 69, n. 1, p. 32–47, 1983.
45, 47, 104
GAO, W.; SHAW, R. H.; PAW, K. T. Observation of organized structure in
turbulent-flow within and above a forest canopy. Boundary-Layer
Meteorology, v. 47, p. 349–377, 1989. 116
123
GAO, Z.; BIAN, L.; ZHOU, X. Measurements of turbulent transfer in the
near-surface layer over a rice paddy in China. Journal of Geophysical
Research - Atmospheres, v. 108, n. D13, p. 4387, 2003. 116
GRECO, A.; MATTHAEUS, W. H.; SERVIDIO, S.; CHUYCHAI, P.; DMITRUK,
P. Statistical analysis of discontinuities in solar wind ACE data and comparison
with intermittent MHD turbulence. The Astrophysical Journal Letters,
v. 691, p. L111–L114, 2009. 88
HADA, T.; KOGA, D.; YAMAMOTO, E. Phase coherence of MHD waves in the
solar wind. Space Science Reviews, v. 107, p. 463–466, 2003. 33, 51, 55, 69, 70,
74, 82, 135
HAMILTON, K.; SMITH, C. W.; VASQUEZ, B. J.; LEAMON, R. J. Anisotropies
and helicities in the solar wind inertial and dissipation ranges at 1 AU. Journal of
Geophysical Research - Space Physics, v. 113, p. A01106, 2008. 54
HARMAN, I. N.; FINNIGAN, J. J. A simple unified theory for flow in the canopy
and roughness sublayer. Boundary-Layer Meteorology, v. 123, p. 339–363,
2007. 116
HE, K. Crisis-induced transition to spatiotemporally chaotic motions. Physical
Review Letters, v. 80, p. 696–699, 1998. 92, 93, 96, 149
HE, K.; CHIAN, A. C.-L. On-off collective imperfect phase synchronization and
bursts in wave energy in a turbulent state. Physical Review Letters, v. 91, p.
034102, 2003. 26, 40, 51, 89, 92, 96, 115, 149
. Critical dynamic events at the crisis of transition to spatiotemporal chaos.
Physical Review E, v. 69, p. 026207, 2004. 92, 96, 149
. Nonlinear dynamics of turbulent waves in fluids and plasmas. Nonlinear
Processes in Geophysics, v. 12, p. 13–24, 2005. 26, 51, 89, 92, 98, 115, 149
HE, K.; SALAT, A. Hysteresis and onset of chaos in periodically driven nonlinear
drift waves. Plasma Physics and Controlled Fusion, v. 31, p. 123–141, 1989.
92, 93, 98, 149
HOF, B.; DE LOZAR, A.; KUIK, D. J.; WESTERWEEL, J. Repeller or attractor?
Selecting the dynamical model for the onset of turbulence in pipe flow. Physical
Review Letters, v. 101, p. 214501, 2008. 117
124
HOF, B.; WESTERWEEL, J.; SCHNEIDER, T. M.; ECKHARDT, B. Finite
lifetime of turbulence in shear flows. Nature, v. 443, p. 59–62, 2006. 117
HOWES, G. G.; DORLAND, W.; COWLEY, S. C.; HAMMETT, G. W.;
QUATAERT, E.; SCHEKOCHIHIN, A. A.; TATSUNO, T. Kinetic simulations of
magnetized turbulence in astrophysical plasmas. Physical Review Letters,
v. 100, p. 065004, 2008. 115
HSU, G.-H.; OTT, E.; GREBOGI, C. Strange saddles and the dimensions of their
invariant manifolds. Physics Letters A, v. 127, p. 199–204, 1988. 36, 37, 98, 100
KAMIDE, Y.; CHIAN, A. C.-L. Handbook of the Solar-Terrestrial
Environment. Berlin: Springer, 2007. 25, 51
KANTZ, H.; GRASSBERGER, P. Repellers, semi-attractors, and long-lived
chaotic transients. Physica D, v. 17, p. 75–86, 1985. 36, 98
KELLOGG, P. J.; HORBURY, T. S. Rapid density fluctuations in the solar wind.
Annales Geophysicae, v. 23, p. 3765–3773, 2005. 53, 56, 87
KOGA, D.; CHIAN, A. C.-L.; HADA, T.; REMPEL, E. L. Experimental evidence
of phase coherence of magnetohydrodynamic turbulence in the solar wind: Geotail
satellite data. Philosophical Transactions of the Royal Society of London -
Mathematical and Physical Sciences, v. 366, p. 447–457, 2008. 33, 51, 55, 69,
70, 82
KOGA, D.; CHIAN, A. C.-L.; MIRANDA, R. A.; REMPEL, E. L. Intermittent
nature of solar wind turbulence near the Earth’s bow shock: phase coherence and
non-Gaussianity. Physical Review E, v. 75, p. 046401, 2007. 33, 51, 55, 69, 70,
72, 74, 82, 135
KOGA, D.; HADA, T. Phase coherence of foreshock MHD waves: wavelet analysis.
Space Science Review, v. 107, p. 495–498, 2003. 33, 34, 51, 55, 69, 70
KOLMOGOROV, A. N. C. R. Acad. Nauk. SSSR, v. 98, p. 527, 1941. 25, 30
KULKARNI, J. R.; SADANI, L. K.; MURTHY, B. S. Wavelet analysis of
intermittent turbulent transport in the atmospheric surface layer over a Monsoon
trough region. Boundary-Layer Meteorology, v. 90, p. 217–239, 1999. 81
125
KURAMOTO, Y. Chemical oscillations, waves and turbulence. Berlin:
Springer-Verlag, 1984. 41
LAI, Y.-C.; BOLLT, E. M.; LIU, Z. Low-dimensional chaos in high-dimensional
phase space: how does it occur? Chaos, Solitons and Fractals, v. 15, p.
219–232, 2003. 45, 47
LAI, Y.-C.; FREI, M. G.; OSORIO, I. Detecting and characterizing phase
synchronization in nonstationary dynamical systems. Physical Review E, v. 73,
p. 026214, 2006. 40
LANGHANS, K.; SCHMIDT, W. Center-to-limb variation of the G-band lines at
430.5 nm. Astronomy and Astrophysics, v. 382, n. 1, p. 312–318, 2001. 135
LEAMON, R. J.; MATTHAEUS, W. H.; SMITH, C. W.; ZANK, G. P.; MULLAN,
D. J.; OUGHTON, S. MHD-driven kinetic dissipation in the solar wind and
corona. The Astrophysical Journal, v. 537, p. 1054–1062, 2000. 60
LUCEK, E. A.; HORBURY, T. S.; BALOGH, A.; DANDOURAS, I.; REME, H.
Cluster observations of structures at quasi-parallel bow shocks. Annales
Geophysicae, v. 22, p. 2309–2314, 2004. 56, 87
MACEK, W. M. Testing for an attractor in the solar wind flow. Physica D,
v. 122, p. 254–264, 1998. 89
MACEK, W. M.; BRUNO, R.; CONSOLINI, G. Generalized dimensions for
fluctuations in the solar wind. Physical Review E, v. 72, p. 017202, 2005. 60
MANSCHER, M.; NORDAHN, M.; MOSEKILDE, E.; MAISTRENKO, L. Riddled
basins of attraction for synchronized type-I intermittency. Physical Letters A,
v. 238, p. 358–364, 1998. 47
MARTIN, D. U.; YUEN, H. C. Spreading of energy in solutions of the nonlinear
Schrodinger equation. Physics of Fluids, v. 23, p. 1269–1271, 1980. 40
MATTHAEWS, W. H.; GOLDSTEIN, M. L.; C., S. Evaluation of magnetic
helicity in homogeneous turbulence. Physical Review Letters, v. 48, p.
1256–1259, 1982. 25
MCCOMAS, D. J.; BAME, S. J.; BARKER, P.; FELDMAN, W. C.; PHILLIPS,
J. L.; RILEY, P.; GRIFFEE, J. W. Solar Wind Electron Proton Alpha Monitor
126
(SWEPAM) for the Advanced Composition Explorer. Space Science Review,
v. 86, p. 563–612, 1998. 60
MIRANDA, R. A.; CHIAN, A. C.-L.; DASSO, S.; ECHER, E.; MUNOZ, P. R.;
TRIVEDI, N. B.; TSURUTANI, B. T.; YAMADA, M. Observation of
non-Gaussianity and phase synchronization in intermittent magnetic field
turbulence in the solar-terrestrial environment. In: INTERNATIONAL
ASTONOMICAL UNION, IAUS264., 2009, Rio de Janeiro, Brazil. Proceedings
IAU Symposium No. 264, 2009. Cambridge: Cambridge University Press,
2009. p. 363–368. ISBN 978-0-521-76499-5. 73, 76, 77
MITIC, C. M.; MASSMAN, W. J.; SCHUEPP, P. H.; COLLETT JR., J. L.
Structural analysis and flux associations of CO2, H2O, heat and ozone over cotton
and grape canopies. Atmospheric Environment, v. 33, p. 1159–1173, 1999. 116
MOLENAAR, D.; CLERCX, H. J. H.; VAN HEIJST, G. J. F. Attractor crisis and
bursting in a fluid flow with two no-slip directions. Physical Review E, v. 75, p.
036309, 2007. 116
MONIN, A. S.; YAGLOM, A. M. Statistical fluid mechanics volume 1. [S.l.]:
Dover Publications, 1971. 30, 32
NARITA, Y.; GLASSMEIER, K.-H.; TREUMANN, R. A. Wave-number spectra
and intermittency in the terrestrial foreshock region. Physical Review Letters,
v. 97, p. 191101, 2006. 53
NARIYUKI, Y.; HADA, T. Remarks on nonlinear relation among phases and
frequencies in modulational instabilities of parallel propagating Alfven waves.
Nonlinear Processes in Geophysics, v. 13, p. 425–441, 2006. 33, 69, 70
NARIYUKI, Y.; HADA, T.; TSUBOUCHI, K. On nonlinear evolution of Alfvenic
turbulence in low beta plasmas. Physics of Plasmas, v. 15, p. 114502, 2008. 33,
69, 70
NICHOLSON, D. R. Introduction to plasma theory. New York: John Wiley &
Sons, 1983. 91
OHKITANI, K.; YAMADA, M. Temporal intermittency in the energy cascade
process and local Lyapunov analysis in fully-developed model turbulence.
Progress of Theoretical Physics, v. 81, p. 329–341, 1989. 40, 111
127
OTT, E. Chaos in dynamical systems. Cambridge: Cambridge University
Press, 1993. 37, 43
OUELLETTE, N. T.; GOLLUB, J. P. Dynamic topology in spatiotemporal chaos.
Physics of Fluids, v. 20, p. 064104, 2008. 36
PALLOCCHIA, G.; BRUNO, R.; BAVASSANO, B.; et al. Turbulence in the solar
wind as seen by Cluster CIS experiment: Preliminary results on intermittency and
scaling laws. In: SOLSPA 2001, Proceedings of the Second Solar Cycle
and Space Weather Euroconference. [S.l.]: ESA Special Publication, 2002. p.
361–364. 53
PAPOULIS, A. Probability, random variables, and stochastic processes.
New York: McGraw-Hill, 1965. 65
PARKER, T. S.; CHUA, L. O. Practical numerical algorithms for chaotic
systems. New York: Springer-Verlag, 1989. 42, 43
PARKS, G. K.; LEE, E.; MOZER, F.; WILBER, M.; LUCEK, E.; DANDOURAS,
I.; REME, H.; MAZELLE, C.; CAO, J. B.; MEZIANE, K.; GOLDSTEIN, M. L.;
ESCOUBET, P. Larmor radius size density holes discovered in the solar wind
upstream of Earth’s bow shock. Physics of Plasmas, v. 13, p. 050701, 2006. 87
PASCHMANN, G.; DALY, P. W. Analysis methods for multi-spacecraft
data. The Netherlands: ESA Publications Division, 2000. 60, 61
PEIXINHO, J.; MULLIN, T. Decay of turbulence in pipe flow. Physical Review
Letters, v. 96, p. 094501, 2006. 117
PEREGRINE, D. H. Calculations of the development of an undular bore. Journal
of Fluid Mechanics, v. 25, p. 321–330, 1966. 92
PETZOLD, L. R. Automatic selection of methods for solving stiff and nonstiff
systems of ordinary differential equations. SIAM Journal of Scientific
Statistics and Computation, v. 4, p. 136–148, 1983. 94
PLATT, N.; SPIEGEL, E. A.; TRESSER, C. On-off intermittency: a mechanism
for bursting. Physical Review Letters, v. 70, n. 3, p. 279–282, 1993. 47
POLITANO, H.; POUQUET, A. von-Karman-Howarth equation for
magnetohydrodynamics and its consequences on third-order longitudinal structure
and correlation functions. Physical Review E, v. 57, p. R21–R24, 1998. 142, 143
128
POLYGIANNAKIS, J. M.; MOUSSAS, X. Detection of nonlinear dynamics in
solar wind and a comet using phase-correlation measures. Solar Physics, v. 158,
p. 159–172, 1995. 40
POWELL, G. E.; PERCIVAL, I. C. A spectral entropy method for distinguishing
regular and irregular motion of hamiltonian systems. Journal of Physics A:
Mathematical and General, v. 12, p. 2053–2071, 1979. 39, 96
QIU, J.; GU, Z. L.; WANG, Z. S. Numerical study of the response of an
atmospheric surface layer to a spatially nonuniform plant canopy.
Boundary-Layer Meteorology, v. 127, p. 293–311, 2008. 116
RAMOS, F. M.; BOLZAN, M. J. A.; SA, L. D. A.; ROSA, R. R. Atmospheric
turbulence within and above an Amazon forest. Physica D, v. 193, p. 278–291,
2004. 81
RAUPACH, M. R.; FINNIGAN, J. J.; BRUNET, Y. Coherent eddies and
turbulence in vegetation canopies: the mixing-layer analogy. Boundary-Layer
Meteorology, v. 78, p. 351–382, 1996. 116
REIDENBACH, M. A.; KOSEFF, J. R.; MONISMITH, S. G. Laboratory
experiments of fine-scale mixing and mass transport within a coral canopy.
Physics of Fluids, v. 19, p. 075107, 2007. 116
REME, H.; AOUSTIN, C.; et al. First multispacecraft ion measurements in and
near the Earth’s magnetosphere with the identical Cluster Ion Spectrometry (CIS)
experiment. Annales Geophysicae, v. 19, p. 1303–1354, 2001. 60
REMPEL, E. L.; CHIAN, A. C.-L. Origin of transient and intermittent dynamics
in spatiotemporal chaotic systems. Physical Review Letters, v. 98, p. 014101,
2007. 37, 89, 92, 93, 100, 102, 103, 149
REMPEL, E. L.; CHIAN, A. C.-L.; MACAU, E. E. N.; ROSA, R. R. Analysis of
chaotic saddles in low-dimensional dynamical systems: the derivative nonlinear
Schrodinger equation. Physica D, v. 199, p. 407–424, 2004. 36, 37
REMPEL, E. L.; CHIAN, A. C.-L.; MIRANDA, R. A. Chaotic saddles at the onset
of intermittent spatiotemporal chaos. Physical Review E, v. 76, p. 056217, 2007.
37, 89, 93, 96, 98, 100, 103, 116
129
REMPEL, E. L.; MIRANDA, R. A.; CHIAN, A. C.-L. Spatiotemporal
intermittency and chaotic saddles in the regularized long-wave equation. Physics
of Fluids, v. 21, p. 074105, 2009. 95, 96, 97, 99, 101, 102, 103
SAHRAOUI, F. Diagnosis of magnetic structures and intermittency in
space-plasma turbulence using the technique of surrogate data. Physical Review
E, v. 78, p. 026402, 2008. 33, 51, 55, 69, 70
SALMOND, J. A.; OKE, T. R.; GRIMMOND, C. S. B.; ROBERTS, S.;
OFFERLE, B. Venting of heat and carbon dioxide from urban canyons at night.
Journal of Applied Meteorology, v. 44, p. 1180–1194, 2005. 116
SAMSONOV, A. A.; ALEXANDROVA, O.; LACOMBE, C.; MAKSIMOVIC, M.;
GARY, S. P. Proton temperature anisotropy in the magnetosheath: comparison of
3-D MHD modelling with Cluster data. Annales Geophysicae, v. 25, p.
1157–1173, 2007. 63
SCHWARTZ, S. J. Shocks: commonalities in solar-terrestrial chains. Space
Science Review, v. 124, p. 333–344, 2006. 87
SHANNON, C. E. A mathematical theory of communication. Bell System
Technical Journal, v. 27, p. 379–423, 1949. 39, 145
SHIMADA, I.; NAGASHIMA, T. A numerical approach to ergodic problem of
dissipative dynamical systems. Progress of Theoretical Physics, v. 61, p.
1605–1616, 1979. 43, 108
SILVA DIAS, M. A. F.; RUTLEDGE, S.; KABAT, P.; DIAS, P. L. S.; NOBRE, C.;
FISCH, G.; DOLMAN, A. J.; ZIPSER, E.; GARSTANG, M.; MANZI, A. O.;
FUENTES, J. D.; ROCHA, H. R.; MARENGO, J.; PLANA-FATTORI, A.; SA, L.
D. A.; ALVALA, R. C. S.; ANDREAE, M. O.; ARTAXO, P.; GIELOW, R.;
GATTI, L. Cloud and rain processes in a biosphere-atmosphere interaction context
in the Amazon region. Journal of Geophysical Research - Atmospheres,
v. 107, n. D20, p. 8072, 2002. 79
SMITH, C. W.; HAMILTON, K.; VASQUEZ, B. J.; LEAMON, R. J. Dependence
of the dissipation range spectrum of interplanetary magnetic fluctuations on the
rate of energy cascade. The Astrophysical Journal Letters, v. 645, p.
L85–L88, 2006. 54
130
SMITH, C. W.; L’HEUREUX, J.; NESS, N. F.; NA, M. H. A.; BURLAGA, L. F.;
SCHEIFELE, J. The ACE magnetic fields experiment. Space Science Review,
v. 86, p. 613–632, 1998. 57
STASIEWICZ, K.; LONGMORE, M.; BUCHERT, S.; SHUKLA, P. K.;
LAVRAUD, B.; PICKETT, J. Properties of fast magnetosonic shocklets at the
bow shock. Geophysical Research Letters, v. 30, n. 24, p. 2241, 2003. 56, 87
STOUGHTON, T. E.; MILLER, D. R.; HUDDLESTON, E. W.; ROSS, J. B.
Evapotranspiration and turbulent transport in an irrigated desert orchard.
Journal of Geophysical Research - Atmosphere, v. 107, p. 4425, 2002. 116
SU, H.-B.; SHAW, R. H.; U, K. T. P.; MOENG, C.-H.; SULLIVAN, P. P.
Turbulent statistics of neutrally stratified flow within and above a sparse forest
from large-eddy simulation and field observations. Boundary-Layer
Meteorology, v. 88, p. 363–397, 1998. 116
SWEET, D.; NUSSE, H. E.; YORKE, J. A. Stagger-and-step method: detecting
and computing chaotic sadddles in higher dimensions. Physical Review Letters,
v. 86, p. 2261–2264, 2001. 37, 38, 107
TAJIMA, T. Computational plasma physics: with applications to fusion
and astrophysics. Massachusetts: Addison-Wesley, 1989. 94
TAN, C.; CHEN, P. F.; ABRAMENKO, V.; WANG, H. Evolution of optical
penumbral and shear flows associated with the X3.4 flare of 2006 december 13.
The Astrophysical Journal, v. 690, p. 1820–1828, 2009. 135
TASS, P.; ROSENBLUM, M. G.; WEULE, J.; KURTHS, J.; PIKOVSKY, A.;
VOLKMANN, J.; SCHNITZLER, A.; FREUND, H.-J. Detection of n:m phase
locking from noisy data: application to magnetoencephalography. Physical
Review Letters, v. 81, p. 3291–3294, 1998. 40
TAYLOR, G. I. The spectrum of turbulence. Proceedings of the Royal Society
of London A, v. 164, p. 476–490, 1938. 31
TEL, T.; LAI, Y.-C. Chaotic transients in spatially extended systems. Physics
Reports, v. 460, p. 245–275, 2008. 36
131
TELLONI, D.; ANTONUCCI, E.; BRUNO, R.; D’AMICIS, R. Persistent and
self-similar large-scale density fluctuations in the solar corona. The
Astrophysical Journal, v. 693, p. 1022–1028, 2009. 33, 51, 69, 70
THYAGARAJA, A. Recurrent motions in certain continuum dynamical systems.
Physics of Fluids, v. 22, p. 2093–2096, 1979. 40
TSUNETA, S.; ICHIMOTO, K.; et al. The solar optical telescope for the Hinode
mission: an overview. Solar Physics, v. 249, n. 2, p. 167–196, 2008. 135
VELTRI, P. MHD turbulence in the solar wind: self-similarity, intermittency and
coherent structures. Plasma Physics and Controlled Fusion, v. 41, p.
A787–A795, 1999. 88
VELTRI, P.; MANGENEY, A. Scaling laws and intermittent structures in solar
wind MHD turbulence. Solar Wind Nine, p. 543–546, 1999. 88
WANG, Y. S.; MILLER, D. R.; ANDERSON, D. E.; CIONCO, R. M.; LIN, J. D.
A spatial length scale analysis of turbulent temperature and velocity fluctuations
within and above an orchard canopy. Boundary-Layer Meteorology, v. 59, p.
125–139, 1992. 116
WELCH, P. The use of fast Fourier transform for the estimation of power spectra:
a method based on time averaging over short, modified periodograms. IEEE
Transactions Audio and Electroacoustics, AU-15, n. 2, p. 70–73, 1967. 60
WESSON, K. H.; KATUL, G. G.; SIQUEIRA, M. Quantifying organization of
atmospheric turbulent eddy motion using nonlinear time series analysis.
Boundary-Layer Meteorology, v. 106, p. 507–525, 2003. 116
WILLIS, A. P.; KERSWELL, R. R. Critical behavior in the relaminarization of
localized turbulence in pipe flow. Physical Review Letters, v. 98, p. 014501,
2007. 117
WOLF, A.; SWIFT, J. B.; SWINNEY, H. L.; VASTANO, J. A. Determining
Lyapunov exponents from a time series. Physica D, v. 16, p. 285–317, 1985. 94
XI, H.; GUNTON, J. D. Spatiotemporal chaos in a model of Rayleigh-Benard
convection. Physical Review E, v. 52, p. 4963–4975, 1995. 39, 96
132
YAMADA, M.; OHKITANI, K. Asymptotic formulas for the Lyapunov spectrum
of fully developed shell model turbulence. Physical Review E, v. 57, p.
R6257–R6260, 1998. 43, 108, 111
YAMADA, M.; OKHITANI, K. Lyapunov spectrum of a model of two-dimensional
turbulence. Physical Review Letters, v. 60, n. 11, p. 983–986, 1988. 108, 110
YAMADA, M.; SAIKI, Y. Chaotic properties of a fully developed model
turbulence. Nonlinear Processes in Geophysics, v. 14, p. 631–640, 2007. 43
YUE, W.; PARLANGE, M. B.; MENEVEAU, C.; ZHU, W.; VAN HOUT, R.;
KATZ, J. Large-eddy simulation of plant canopy flows using plant-scale
representation. Boundary-Layer Meteorology, v. 124, p. 183–203, 2007. 116
ZHOU, Y.; MATTHAEUS, W. H.; DMITRUK, P. Magnetohydrodynamic
turbulence and time scales in astrophysical and space plasmas. Reviews Modern
Physics, v. 76, p. 1015–1035, 2004. 88
ZHU, W.; VAN HOUT, R.; KATZ, J. PIV measurements in the atmospheric
boundary layer within and above a mature corn canopy. Part II: Quadrant-hole
analysis. Journal of Atmospheric Science, v. 64, p. 2825–2838, 2007. 116
133
7 APPENDIX A - SYNCHRONIZATION IN THE SOLAR PHOTO-
SPHERE BEFORE AND AFTER A SOLAR FLARE EVENT
Solar images obtained by the Hinode mission provide continuous high-resolution ob-
servations of the solar photosphere, the solar chromosphere and solar corona. The
Solar Optical Telescope (SOT) (TSUNETA et al., 2008) onboard Hinode provides G-
band (430.5 nm) images with a spatial resolution of 0.109′′ per pixel. The G-band is
a molecular band in the solar spectrum which consists mainly of electronic transi-
tions between rotational and vibrational sublevels of the CH molecule (LANGHANS;
SCHMIDT, 2001). It has been used to investigate photospheric bright points, and as
a tracer to obtain the horizontal component of sunspot penumbral and shear flows
(TAN et al., 2009). Shortly after launch, Hinode observed a X3.4 solar flare in active
region AR 10930 at 02:14 UT on 13 December 2006. Following Tan et al. (2009)
we select two images, one obtained before the solar flare at 01:00:32 UT and one
obtained after the flare at 04:36:37 UT. During this period the projection effects are
minimal (ABRAMENKO et al., 2008). Figure 7.1 shows the G-band image taken before
the flare. A white line separates two selected areas with the same size (772x1020 pix-
els), one containing AR 10930 and the other containing a quiet region. The postflare
G-band image is also separated in two areas with the same size containing the active
and quiet regions.
Figure 7.2 shows kurtosis and the phase coherence index as a function of spatial scale
r computed from the two selected pre-flare regions. We observe that the variation
of kurtosis with r in the quiet region, within the inertial subrange, is close to a
Gaussian process (K = 0), and is scale-invariant for all scales r>∼1 Mm, consistent
with the features of a monofractal process (ABRAMENKO et al., 2002). The active
region, on the other hand, displays an increase of kurtosis as the spatial scale r
decreases until r ∼ 2 Mm, which is a characteristic of a non-Gaussian process related
to nonlinear energy cascade within the inertial subrange, and these values are higher
than those obtained from the quiet region for scales<∼ 7 Mm. The degree of phase
synchronization measured by the phase coherence index (HADA et al., 2003; KOGA
et al., 2007; CHIAN; MIRANDA, 2009) in the active region increases with decreasing
spatial scale r, while the degree of synchronization of the quiet region is lower at all
scales.
The high-resolution solar images from Hinode allows us to quantify the degree of non-
Gaussianity and phase synchronization at smaller spatial scales than those obtained
135
AR 10930 Quiet region
454 504 554 604
40
60
80
100
120
140
S−
N (
arcs
econ
d)
E−W (arcsecond)
Hinode SOT G−band at 01:00:32 UT on 13 December 2006
FIGURE 7.1 - Hinode SOT G-band image taken at 01:00:32 UT on 13 December 2006.The white line separates two areas containing the active region AR 10930,and a quiet region.
from solar images from SOHO. Fig. 7.2 shows that the degree of non-Gaussianity
and synchronization of AR 10930 and the quiet region increases for scales r<∼1 Mm,
which indicates that both regions display intermittent features for r<∼1.
Next we compare the degree of amplitude and phase synchronization of the active
region before and after the flare. Figure 7.3 shows kurtosis and the phase coherence
index as a function of spatial scale r computed from AR 10930 before (continuous
line) and after (dashed line) the solar flare. The degree of intermittency and syn-
chronization in the active region is almost the same before and after the flare, being
only slightly higher at scales r>∼1 before the solar flare. The degree of phase syn-
chronization measured by the phase coherence index is consistent with the variation
of kurtosis.
136
0
1
2
3
4
Kurt
osi
s
AR 10930Quiet region
Before flare
10-1
100
101
r (Mm)
0
0.2
0.4
Phas
e co
her
ence
index
FIGURE 7.2 - Kurtosis (upper panel) and the phase coherence index (lower panel) as afunction of spatial scale r computed from AR 10930 (black line) and thequiet region (grey line) before the solar flare.
0
1
2
3
4
Kurt
osi
s
Before flare After flare
AR 10930
10-1
100
101
r (Mm)
0
0.2
0.4
Phas
e co
her
ence
index
FIGURE 7.3 - Kurtosis (upper panel) and the phase coherence index (lower panel) as afunction of spatial scale r computed from AR 10930 before (black line) andafter (dashed line) the solar flare.
137
8 APPENDIX B - KOLMOGOROV 1941 THEORY AND ITS EXTEN-
SION TO MAGNETOHYDRODYNAMICS
8.1 Neutral fluids
The dynamics of incompressible, neutral fluids can be described by the Navier-Stokes
equations
∂tu + (u · ∇)u = −∇p+ ν∆u + f , (8.1)
∇ · u = 0. (8.2)
where u = u(x, t) denotes the fluid velocity which depends on position x and time
t, p is the pressure, ν represents the kinematic viscosity, and f is an external force.
Let us define u1 as the i-th component of the fluid velocity at point x1 and u2 as
the j-th component of the velocity at point x2 separated from point x1 by a distance
|r| = r. Rewriting Equation (8.1)
∂tu1i + ∂1ku1ku1i = −∂1ip+ ν∆1u1i + f1i, (8.3)
∂tu2j + ∂2ku2ku2j = −∂2jp+ ν∆1u2j + f2j, (8.4)
where we have made use of Einstein’s notation. Next, we multiply Equation (8.3) by
u2j and equation (8.4) by u1i, sum the two resulting equations and take the average
∂t 〈u1iu2j〉+ ∂1k 〈u2ju1ku1i〉+ ∂2k 〈u1iu2ku2j〉
= −∇1i 〈u2jp〉 − ∇2j 〈u1ip〉+ ν∆1 〈u2ju1i〉+ ν∆2 〈u1iu2j〉
+ 〈u1if2j〉+ 〈u2jf1i〉 , (8.5)
where the 〈 〉 denote the ensemble average. Let us define the correlation functions
139
〈u1iu2j〉 = Cij(r), (8.6)
〈u1iu1ku2j〉 = Cik,j(r), (8.7)
and partial derivatives
∂1k = −∂rk , (8.8)
∂2k = ∂rk . (8.9)
Equation (8.5) can be written as
∂tCij(r) = ∂rkCik,j(r) + ∂rkCjk,i(r) + 2ν∆Cij(r) + 2εij(r), (8.10)
where εij is a tensor related to energy dissipation rate. Taking the trace of equation
(8.10) and integrating with respect to rk one obtains
Cik,i(r) + ν∂rkCii(r) = −1
3εrk. (8.11)
where ε is the mean energy dissipation rate. Let us define the second and third-order
two-point differences as
S2(r) =⟨(δu1)
2⟩,
S3(r) =⟨(δu1)
3⟩,
where δu1(r) = u1(x1 + r)− u1(x1). Assuming homogeneity and isotropy, the corre-
lation functions are related to two-point differences (or structure functions) by the
following identities
140
Cii(r) =1
2r2∂rr
3S2(r), (8.12)
Cik,i(r) =1
12
rkr3∂rr
4S3(r). (8.13)
Inserting (8.12) and (8.13) into (8.11) one obtains
S3(r)− ν∂rS2(r) = −4
5εr. (8.14)
In the limit ν → 0
S3(r) = −4
5εr. (8.15)
where r = |r| represents spatial scale. Equation (8.15) was obtained by Kolmogorov
in 1941. By assuming that turbulence is self-similar at small scales, i. e. it posesses
a unique scaling exponent
δu(λr) = λαδu(r). (8.16)
Subtituting (8.16) into (8.15)
λ3αS3 = −4
5ελr, (8.17)
hence the scaling exponent should be equal to 1/3.
Eq. (8.15) can be generalized to structure functions of order p
Sp(r) = 〈(δu)p〉 . (8.18)
From the self-similarity assumption we can infer that, if S3 ∝ r for p = 3, then in
general Sp ∝ rαp = rp/3, and since (εr)p/3 has exactly the same dimensions as Sp for
141
p = 3, the structure function of order p should obey (FRISCH, 1995)
Sp = Cpεp/3rp/3, (8.19)
where Cp is a dimensionless constant. For p = 3, Cp = −4/5.
The second-order structure function is related with the energy spectrum by (DAVID-
SON, 2004)
⟨(δu(r))2⟩ ∼ ∫ ∞
π/r
E(k)dk, (8.20)
where E(k) represents the energy of eddies of size r ∼ π/k. Combining Eq. (8.20)
with Eq. (8.19) with p = 2, and taking the derivative with respect to k one can
obtain
E(k) ∼ C ′2ε2/3k−5/3, (8.21)
where C ′2 = −(2C2π2/3)/3.
8.2 Magnetized flows
A similar derivation for turbulent magnetized flows was carried out by Politano and
Pouquet (1998). The magnetohydrodinamical (MHD) equations which govern the
dynamics of incompressible conducting flows can be written in terms of Elsasser
variables as
(∂t + z∓ · ∇
)z± = −∇P∗ + ν+∇2z± + ν−∇2z∓, (8.22)
∇ · u = 0, (8.23)
∇ · b = 0, (8.24)
where z± = u ± b represent the Elsasser variable, b is the magnetic field, P∗ =
P + |b|2/2 is the total pressure, and ν± = (ν ± η)/2 where ν is the viscosity and η
142
is the magnetic diffusivity.
Asumming homogeneity and isotropy, and considering only the longitudinal com-
ponent of the Elsasser variables (i.e., the components parallel to the displacement
vector r), Politano and Pouquet (1998) obtained the following relation
⟨(δz+
L (r))2δz−L (r)
⟩− 2
⟨z+L (x)z+
L (x)z−L (x′)⟩
= −Cdε+r (8.25)⟨(δz−L (r)
)2δz+L (r)
⟩− 2
⟨z−L (x)z−L (x)z+
L (x′)⟩
= −Cdε−r (8.26)
where the subscript L denotes longitudinal components, δz±L = z±L (x′) − z±L (x) are
the two-point differences of the longitudinal components of the Elsasser variables z±L ,
x′ = x+r, ε± denote the dissipation rate of 〈|z±|2〉 /2 respectively, and Cd = 2Kd/3,
K3 = 4/5. Equations (8.25) and (8.26) can be written in terms of the original
physical variables of MHD, namely, the velocity and the magnetic field
⟨δu3
L(r)⟩− 6
⟨b2L(x)uL(x + r)
⟩= −Kdε
T r (8.27)
−⟨δb3L(r)
⟩+ 6
⟨u2L(x)bL(x + r)
⟩= −Kdε
Cr (8.28)
where 2εT = ε+ + ε−, 2εC = ε+ − ε−. Kolmogorov’s result is recovered for b = 0.
Note that Equations (8.25) and (8.26) couple the third-order structure function to
the third-order correlation function, hence they do not directly provide a scaling
law for the third-order structure function (BISKAMP, 2003). Eqs. (8.25) and (8.26)
can be simplified when the ratio 〈|b|2〉 / 〈|u|2〉 ∼ 1. In this case the z± fields evolve
quasi-independently, and one can assume that they are not correlated at different
spatial locations. Hence, 〈b2L(x)uL(x + r)〉 = 〈u2L(x)bL(x + r)〉 = 0, and one obtains
(POLITANO; POUQUET, 1998)
⟨(δz+
L (r))2δz−L (r)
⟩= −Cdε+r, (8.29)⟨(
δz−L (r))2δz+
L (r)⟩
= −Cdε+r, (8.30)
143
which are closer to Kolmogorov’s law shown in Eq. (8.15). In terms of velocity and
magnetic fields, Eqs. (8.29) and (8.30) read
⟨δu3
L(r)⟩−⟨δuL(r)δb2L(r)
⟩= −Kdε
T r, (8.31)
−⟨δb3L(r)
⟩+⟨δbL(r)δu2
L(r)⟩
= −KdεT r, (8.32)
144
9 APPENDIX C - SHANNON ENTROPY
In this Appendix we derive the Shannon entropy following the original derivation
by Shannon (1949). Let H = H(p1, p2, ..., pn) denote the “amount of uncertainty”
(entropy) of a specific process. We make the following assumptions:
a)H should be continuous in pi.
b)If all pi are equal, pi = 1/N , then H should be a monotonic increasing func-
tion of N (with equally likely events there is more choice, or uncertainty,
when there are more possible events).
c)If a choice is decomposed into two successive choices, then the original H
should be the weighted sum of the individual values of H. For example,
in Figure 9.1, we can decompose three possibilities p1 = 1/2, p2 = 1/3
and p3 = 1/6 into two possibilities with probability 1/2, and if the second
occurs then there is another choice with probabilities 2/3 and 1/3. We can
denote this as
H
(1
2,1
3,1
6
)= H
(1
2,1
2
)+
1
2H
(2
3,1
2
)(9.1)
Note that the second term in the right-hand side of Equation (9.1) has a coefficient
1/2, which must be introduced because this second choice only occurs half the time.
1/21/2
1/3
1/61/3
2/3
1/2
FIGURE 9.1 - Decomposition of three possibilities p1 = 1/2, p2 = 1/3 and p3 = 1/6 intotwo possibilities with probability 1/2. If the second occurs then there isanother choice with probabilities 2/3 and 1/3.
SOURCE: Adapted from Shannon (1949)
145
Let us derive first an expression for an uniform distribution. Let
H(1/N, 1/N, ..., 1/N) = A(N). From assumption (c) we can decompose a
choice from sM equally-likely possibilities into a series of M choices from s
equally-likely possibilities. Let us denote
A(sM) = MA(s) (9.2)
A(tN) = NA(t) (9.3)
Let us choose N arbitrarily large and find M which satisfies
sM ≤ tN ≤ sM+1 (9.4)
Taking the logarithm of (9.4) and dividing by N log(s)
log(sM) ≤ log(tN) ≤ log(sM+1)
M log(s) ≤ N log(t) ≤ (M + 1) log(s)
M
N≤ log(t)
log(s)≤ M
N+
1
N(9.5)
Now, since A is a monotonic function:
A(sM) ≤ A(tN) ≤ A(sM+1) (9.6)
From Eq. (9.2) and (9.3)
MA(s) ≤ NA(t) ≤ (M + 1)A(s)
dividing by NA(s) we obtain
146
M
N≤ A(t)
A(s)≤ M
N+
1
N(9.7)
Next, we subtract Eq. (9.8) from Eq. (9.7). Let us rewrite Eq. (9.8)
− M
N≥ − log(t)
log(s)≥ −M
N− 1
N
−MN− 1
N≤ − log(t)
log(s)≤ −M
N(9.8)
Now we can add Eqs. (9.7) and (9.8)
− 1
N≤ A(t)
A(s)− log(t)
log(s)≤ 1
N∣∣∣A(t)A(s)− log(t)
log(s)
∣∣∣ ≤ 1
N.
Taking the limit N →∞
A(t)
A(s)− log(t)
log(s)= 0
A(t)
A(s)=
log(t)
log(s)
A(t) =A(s)
log(s)log(t) (9.9)
where A(s)/ log(s) is a constant which can be absorbed by the base of log(t)
A(t) = log(t) (9.10)
For a non-uniform distribution we can write the probability of each choice pi as
147
pi =niN, ni, N ∈ N (9.11)
The uncertainty of the complete set of outcomes is
A(N) = log(N) (9.12)
Using assumption (c) we can write
log(N) = H(p1, ..., pN) +∑i
piH(ni)
= H(p1, ..., pN) +∑i
pi log(ni)
(9.13)
Solving for H we obtain
H = log(N)−∑i
pi log(ni)
= −(∑i
pi log(ni)− log(N))
= −∑i
pi log(niN
)= −
∑i
pi log pi (9.14)
148
10 APPENDIX D - WAVE ENERGY IN THE BENJAMIN-BONA-
MAHONY EQUATION
The driven-damped regularized long wave equation (Benjamin-Bona-Mahony equa-
tion) is given by (HE; SALAT, 1989; REMPEL; CHIAN, 2007; HE, 1998; HE; CHIAN,
2004; HE; CHIAN, 2003; HE; CHIAN, 2005)
∂tu+ a∂txxu+ c∂xu+ fu∂xu = −νu− ε sin(κx− Ωt), (10.1)
where a, c, and f are constants, ν is a damping parameter, ε is the driver amplitude,
κ is the driver wavenumber and Ω is the driver frequency.
Ignoring the damping and external force we have (BENJAMIN et al., 1972; BENJAMIN,
1972)
∂tu+ a∂txxu+ c∂xu+ fu∂xu = 0. (10.2)
Multiplying Eq. (10.2) by u
u∂tu+ au∂txxu+ cu∂xu+ fu2∂xu = 0. (10.3)
Integrating Eq. (10.3) between x = −∞ and x =∞
∫ ∞−∞
u∂tudx+ a
∫ ∞−∞
u∂txxudx+ c
∫ ∞−∞
u∂xudx+ f
∫ ∞−∞
u2∂xudx = 0. (10.4)
The first term on the left-hand side of Eq. (10.4), after integrating by parts, can be
written as
∫u∂tudx =
1
2
∫∂tu
2dx (10.5)
The second term on the left-hand side of Eq. (10.4) can be written as
149
∫u∂x (∂x∂tu) dx = u (∂x∂tu)−
∫(∂xu) (∂x∂tu) dx (10.6)
The third term on the left-hand side of Eq. (10.4) can be written as
c
∫u∂xudx =
c
2
∫∂xu
2dx
=c
2u2 (10.7)
The fourth term on the left-hand side of Eq. (10.4) can be written as
f
∫u2∂xudx = f
(u3 − 2
∫u2∂xudx
)3f
∫u2∂xudx = fu3
f
∫u2∂xudx = f
u3
3(10.8)
Subtituting Eqs. (10.5), (10.6), (10.7) and (10.8) into Eq. (10.4)
1
2
∫ ∞−∞
∂tu2dx+ au∂x∂tu|∞−∞ − a
∫ ∞−∞
(∂xu) (∂x∂tu) dx+c
2u2∣∣∞−∞ + f
u3
3
∣∣∣∣∞−∞
= 0.
(10.9)
On the assumption that u, ∂xu and ∂x∂tu vanish as x → ±∞ (BENJAMIN, 1972),
Eq. (10.9) gives
150
1
2
∫ ∞−∞
∂tu2dx− a
∫ ∞−∞
(∂xu) [∂t (∂xu)] dx = 0
1
2
∫ ∞−∞
∂tu2dx− a
2
∫ ∞−∞
∂t (∂xu)2 dx = 0
∂t
[1
2
∫ ∞−∞
(u2 − a (∂xu)2) dx] = 0
1
2
∫ ∞−∞
[u2 − a (∂xu)2] dx = Cte (10.10)
Eq. (10.10) defines an invariant quantity of Eq. (10.2), the “wave energy” E (BEN-
JAMIN et al., 1972; BENJAMIN, 1972)
E =1
2
∫ ∞−∞
[u2 − a (∂xu)2] dx (10.11)
151
11 LIST OF PUBLICATIONS
•Koga, D., Chian, A. C.-L., Miranda, R. A., and Rempel, E. L. Intermit-
tent nature of solar wind turbulence near the Earth’s bow shock: phase
coherence and non-Gaussianity. Physical Review E, 75, 046401, 2007.
•Chian, A. C.-L. and Miranda, R. A.. Cluster and ACE observations of
phase synchronization in intermittent magnetic field turbulence: a compar-
ative study of shocked and unshocked solar wind. Annales Geophysicae, 27,
1789-1801, 2009.
•Miranda, R. A., Chian, A. C.-L., Dasso, S., Echer, E., Munoz, P. R.,
Trivedi, N. B., Tsurutani, B. T. and Yamada, M. Observation of non-
Gaussianity and phase synchronization in intermittent magnetic field tur-
bulence in the solar-terrestrial environment. Proceedings IAU symposium
No. 264, 363-368, 2010.
•Chian, A. C.-L., Miranda, R. A., Koga, D., Bolzan, M. J. A., Ramos, F.
M., and Rempel, E. L. Analysis of phase coherence in fully developed at-
mospheric turbulence: Amazon forest canopy. Nonlinear Processes in Geo-
physics, 15, 567-573, 2008.
•Rempel, E. L., Miranda, R. A., and Chian, A. C.-L. Spatiotemporal
intermittency and chaotic saddles in the regularized long-wave equation.
Physics of Fluids, 21, 074105, 2009.
•Chian, A. C.-L., Miranda, R. A., Rempel, E. L., Saiki, Y. and Yamada,
M. Amplitude-phase synchronization in chaotic saddles at the onset of per-
manent spatiotemporal chaos. Physics Review Letters, 2010 (submitted).
153
PUBLICACOES TECNICO-CIENTIFICAS EDITADAS PELO INPE
Teses e Dissertacoes (TDI) Manuais Tecnicos (MAN)
Teses e Dissertacoes apresentadas nosCursos de Pos-Graduacao do INPE.
Sao publicacoes de carater tecnico queincluem normas, procedimentos, in-strucoes e orientacoes.
Notas Tecnico-Cientıficas (NTC) Relatorios de Pesquisa (RPQ)
Incluem resultados preliminares depesquisa, descricao de equipamentos,descricao e ou documentacao de progra-mas de computador, descricao de sis-temas e experimentos, apresentacao detestes, dados, atlas, e documentacao deprojetos de engenharia.
Reportam resultados ou progressos depesquisas tanto de natureza tecnicaquanto cientıfica, cujo nıvel seja com-patıvel com o de uma publicacao emperiodico nacional ou internacional.
Propostas e Relatorios de Projetos(PRP)
Publicacoes Didaticas (PUD)
Sao propostas de projetos tecnico-cientıficos e relatorios de acompan-hamento de projetos, atividades e con-venios.
Incluem apostilas, notas de aula e man-uais didaticos.
Publicacoes Seriadas Programas de Computador (PDC)
Sao os seriados tecnico-cientıficos: bo-letins, periodicos, anuarios e anais deeventos (simposios e congressos). Con-stam destas publicacoes o InternacionalStandard Serial Number (ISSN), que eum codigo unico e definitivo para iden-tificacao de tıtulos de seriados.
Sao a sequencia de instrucoes ou codi-gos, expressos em uma linguagem deprogramacao compilada ou interpre-tada, a ser executada por um computa-dor para alcancar um determinado obje-tivo. Aceitam-se tanto programas fontequanto os executaveis.
Pre-publicacoes (PRE)
Todos os artigos publicados em periodi-cos, anais e como capıtulos de livros.