Б. В. Козелов Полярный геофизический институт КНЦ РАН

24
Суббуревая и буревая активность магнитосферы как отклик на стохастическое воздействие солнечного ветра Б. В. Козелов Полярный геофизический институт КНЦ РАН

Upload: maxima

Post on 09-Jan-2016

47 views

Category:

Documents


0 download

DESCRIPTION

Б. В. Козелов Полярный геофизический институт КНЦ РАН. Суббуревая и буревая активность магнитосферы как отклик на стохастическое воздействие солнечного ветра. Approach. INPUT. “BLACK BOX”. OUTPUT. Activity indexes. Solar wind. Magneto-sphere. B z (t) , v(t)B z (t). AE(t) , Dst(t). - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Б. В. Козелов  Полярный геофизический институт КНЦ РАН

Суббуревая и буревая активность магнитосферы как

отклик на стохастическое воздействие солнечного

ветра

Б. В. Козелов

Полярный геофизический институт КНЦ РАН

Page 2: Б. В. Козелов  Полярный геофизический институт КНЦ РАН

Approach

Magneto-sphere

Solar windActivity indexes

Bz(t), v(t)Bz(t) AE(t), Dst(t)

INPUT “BLACK BOX” OUTPUT

Page 3: Б. В. Козелов  Полярный геофизический институт КНЦ РАН

Stochastic differential equation

From the SDE we can extract the stochastic noise process,

dB(t) dX(t) f [X(t)] dt

g[X(t)]

by computing the drift and diffusion terms from the time series:

f [X] dX X

dt, D[X] g2[X]

dX f [X]dt2

X

dB2

dX(t) f [X(t)]dt g[X(t)]dB(t)

Page 4: Б. В. Козелов  Полярный геофизический институт КНЦ РАН

Stochastic differential equation

It has recently been shown that fluctuations of global quantities (let us name it X(t)) in certain avalanching and turbulent systems can be described by stochastic differential equations (SDEs) driven by a colored noise term with a diffusion coefficient D = g2[X(t)] depending on X(t). The equation also contains a deterministic drift term f [X(t)] which keeps X(t) within certain limits, and the equation takes the form:

Here B(t) could be a Wiener process (Brownian motion), a self-similar Gaussian process (a fractional Brownian motion with Hurst exponent H).

dX(t) f [X(t)]dt g[X(t)]dB(t)

[M. Rypdal, K Rypdal, arXiv:0807.3416v1, 2008]

Page 5: Б. В. Козелов  Полярный геофизический институт КНЦ РАН

Strongly driven Zhang sandpile

[M. Rypdal, K Rypdal, arXiv:0807.3416v1, 2008]

Part of a time series for the toppling activity. b) The increments of the signal. c) The conditional variance of the increments given the value of x. d) The conditional mean of increments.

f [x)]

D[x)]

H=0.75

Page 6: Б. В. Козелов  Полярный геофизический институт КНЦ РАН

2D turbulence simulation

a) Part of a time series for the kinetic energy. b) The increments of the signal. c) The conditional variance of the increments given the value of X. The inset shows the logarithm of this variance versus X. d) The conditional mean of increments.

[M. Rypdal, K Rypdal, arXiv:0807.3416v1, 2008]

D[x)]

f [x)]

Page 7: Б. В. Козелов  Полярный геофизический институт КНЦ РАН

AE data preparation

AE AE/t|t=1 min

Compute the drift and diffusion term

Removing drift term

Normalized by conditional std

Compare the rest noise and SW driver

Page 8: Б. В. Козелов  Полярный геофизический институт КНЦ РАН

Stochastic noise process dB(t) extracted from AE-index 1-min

data

1 year

6 years

f[X]

D[X]

Bursts, intermittency => multifractal structure

Yearly variations, more visible for years of less activity

Page 9: Б. В. Козелов  Полярный геофизический институт КНЦ РАН

Application to magnetosphere dynamics

Magnetosphere dynamics and solar wind beyond second order statistics.[…..]

Generalization from self-similar to MF noise is needed.

Page 10: Б. В. Козелов  Полярный геофизический институт КНЦ РАН

SW data preparation

“Transplantation” algorithm has been used to fill data gaps in Bz(t) and v(t).

ttp tn

Xp Xn

1-min OMNI data

Page 11: Б. В. Козелов  Полярный геофизический институт КНЦ РАН

Usual problems with MF approach to data set

1. Definition of a measure2. Quality of estimated features (error bars?)

Page 12: Б. В. Козелов  Полярный геофизический институт КНЦ РАН

“French” MF approach

Authors: S.Jaffard, B. Lashermes, P. Abry, H. Wendt

~ 10 papers

Strong mathematical background, detailed description of algorithms, error bars for all quantities.

Page 13: Б. В. Козелов  Полярный геофизический институт КНЦ РАН

ReferencesLashermes B., S.G. Roux, P. Abry, and S. Jaffard, Comprehensive multifractal analysis

of turbulent velocity using the wavelet leaders, Eur. Phys. J. B 61, 201–215, 2008.

Wendt H., P.Abry, Bootstrap tests for the time constancy of multifractal attributes, 2007.

Wendt H., P.Abry, and S.Jaffard, Bootstrap for Empirical Multifractal Analysis (Bootstrap application to hydrodynamic turbulence), IEEE SIGNAL PROCESSING MAGAZINE, P.38-48, JULY 2007.

Jaffard S., “Multifractal formalism for functions, part 2: Selfsimilar functions,” SIAM J. of Math. Anal., vol. 28, no. 4, pp. 971–998, 1997.

Jaffard S., B. Lashermes, and P. Abry, “Wavelet leaders in multifractal analysis,” in Wavelet Analysis and Applications, T. Qian, M.I. Vai, X. Yuesheng, Eds. Cambridge, MA: Birkhäuser, 2006, pp. 219–264.

Abry P., P. Flandrin, M. Taqqu, and D. Veitch, “Wavelets for the analysis, estimation and synthesis of scaling data,” in Selfsimilar Network Traffic and Performance Evaluation. Spring 2000, Wiley.

Jaffard S. , B. Lashermes and P. Abry, Wavelet leaders in multifractal analysis, in Wavelet Analysis and Applications, 2005, University of Macau, China.

Efron B., The Jackknife, the Bootstrap, and Other Resampling Plans, Society for Industrial and Applied Mathematics, Philadelphia, 1982.

Page 14: Б. В. Козелов  Полярный геофизический институт КНЦ РАН

Code overview

Discrete wavelet decomposition, Daubechies wavelets.Wavelet leaders (WL) used instead of wavelet coefficients.All MF attributes calculated from WL statistics.Block bootstrap procedure applied at each scale to estimate error bars.Double bootstrap procedure for estimation of temporal stability of the MF characteristic.Simple expansion is possible for forecast.

Page 15: Б. В. Козелов  Полярный геофизический институт КНЦ РАН

Discrete wavelet coefficients (DWC)

Integers j Z and k Z index the scale a = 2j and the location x0 = k2j. Wavelets

{j,k}jZ,kZ are space-shifted and scale-dilated templates of a mother-wavelet 0:

The discrete wavelet transform is a decomposition (also called multiresolution analysis) of the function f on the orthogonal basis {j,k}jZ,kZ composed of discrete wavelets j,k:

and define a basis distributed according to a dyadic basis in the space-scale plane. Every wavelet j,k and then every DWC d(j,k) can be associated to the dyadic interval

(j,k) = [2jk, 2j(k + 1)[ which will be usefully used for indexing the DWC: d(j,k) = d().

Page 16: Б. В. Козелов  Полярный геофизический институт КНЦ РАН

Definition of Wavelet Leaders

[H.Wendt, P. Abry, 2007]

Page 17: Б. В. Козелов  Полярный геофизический институт КНЦ РАН

MF attributesLX(j,k) => SX(j,q) j (q) => (q)

singularity spectrum D(h)

Log-cumulants. This method provides estimates of the parameters cp

of the Taylor series expansion of (q) for q 0 :

If the process under analysis is monofractal then c1 = H 0 and cp = 0 for p > 1.

A non-zero value for c2 explicitly establishes the multifractal (vs. monofractal)

nature of the data and the parameter c2 (also called the intermittency coefficient)

is used to characterize the degree of multifractality. A quadratic approximation of the scaling exponents: (q) c1q + c2q2/2 (when q 0) corresponds to a

quadratic approximation of the singularity spectrum: D(h) 1 + (h−c1)2/2c2 (when

h c1)

Page 18: Б. В. Козелов  Полярный геофизический институт КНЦ РАН

[H.Wendt, P. Abry, 2007] [B. Efron, 1982]

Page 19: Б. В. Козелов  Полярный геофизический институт КНЦ РАН

Time constancy of multifractal attributes

Procedure for obtaining T (left) and T *(right) from the wavelet Leaders {LX(j,k)}

of X. "cut", "estimate" and “" stand for cutting a set into M subsets, computing estimates ˆ, and bootstrap resampling. [H.Wendt, P.Abry, 2007]

Page 20: Б. В. Козелов  Полярный геофизический институт КНЦ РАН

Example of MF features for AE-index

(2000 year)ζ(q) D(h) c1 c2 c3

ζ(q) is estimated from linear regression of log Z(q,a) versus log a on the time scales 4-128 minutes. A bootstrapping technique has been employed that allows estimation of, and reduction of, error bars.

Page 21: Б. В. Козелов  Полярный геофизический институт КНЦ РАН

Example of MF features of vBz (2000 year)

ζ(q) D(h) c1 c2 c3

ζ(q) is estimated from linear regression of log Z(q,a) versus log a on the time scales 4-128 minutes. A bootstrapping technique has been employed that allows estimation of, and reduction of, error bars.

Page 22: Б. В. Козелов  Полярный геофизический институт КНЦ РАН

AE vBz

c1

c2

c3

2000 2001 2002 2003 2004 2005 2000 2001 2002 2003 2004 2005

Page 23: Б. В. Козелов  Полярный геофизический институт КНЦ РАН

Example for Dst – index (1-hour data)

D[X]

f[X]

c1

c2

c3

Page 24: Б. В. Козелов  Полярный геофизический институт КНЦ РАН

Conclusions•The noise process dB(t) extracted from both AE-index and solar wind E-field time series exhibit intermittency characteristics.• At the substorm temporal scales (4-128 min) the leading Hölder exponent c1 (the Hurst exponent) is ~0.40 in AE-index

and ~0.46 in solar wind.• Solar wind is more intermittent than the noise process extracted from AE (|c2| is larger in solar wind time series).• In average, the skewness (c3) of the singularity spectrum (if it

is significant?) have opposite signs in AE-index and solar wind.• There is a seasonal variation in the AE-index features.• If a synthetic multifractal process can be constructed that reflects these characteristics, the stochastic equation can provide a model for these time series.