Analysis of Random Cascades using Wavelet
Analysis: From Theoretical Concepts to Experimental Applications
A. Arneodo
Centre de Recherche Paul Pascal
Avenue Schweitzer, 33600 Pessac, France

 Multiplicative cascade models have enjoyed an increasing
interest in recent years as the paradigm of multifractal objects14.
The notion of cascade actually refers to a selfsimilar process
whose properties are defined multiplicatively from coarse to
fine scales. In that respect, it occupies a central place in
the statistical theory of turbulence3,5. Since Richardson's famous
poem, the turbulent cascade picture has been often invoked to
account for the intermittency phenomenon observed in fully developed
turbulent flows5,6: energy is transfered from large eddies down
to small scales (where it is dissipated) through a cascade process
in which the transfer rate at a given scale is not spatially
homogeneous, as supposed in the theory developed by Kolmogorov7
in 1941, but undergoes local intermittent fluctuations5. Over
the past thirty years, refined models including the lognormal
model of Kolmogorov8 and Obukhov9, multiplicative hierarchical
cascade models like the random betamodel, the alphamodel, the
pmodel (for review, see Ref. [3]), the logstable models10 and
more recently the loginfinitely divisible cascade models1114
with the rather popular logPoisson model advocated by She and
Leveque15, have grown in the literature as reasonable models
to mimic the energy cascading process in turbulent flows. On
a very general ground, a selfsimilar cascade is defined by the
way the scales are refined and by the statistics of the multiplicative
factors at each step of the process3,4,10. One can thus distinguish
discrete cascades that involve discrete scale ratios leading
to logperiodic corrections to scaling (discrete scale invariance16),
from continous cascades without preferable scale factors (continuous
scale invariance). As far as the fragmentation process is concerned,
one can specify whether some conservation laws are operating
or not4; in particular, one can discriminate between conservative
(the measure is conserved at each cascade step) and non conservative
(only some fraction of the measure
 is transfered at each step) cascades. More fundamentally,
there are two main classes of selfsimilar cascade processes:
deterministic cascades that generally correspond to solvable
models and random cascades that are likely to provide more realistic
models but for which some theoretical care is required as far
as their multifractal limit and some basic multifractal properties
(including multifractal phase transitions) are concerned4. As
a notable member of the later class, the independent random cascades
introduced by Mandelbrot6 as a general model of random curdling
in fully developed turbulence, have a special status since they
are the main cascade model for which deep mathematical results
have been obtained17,18. More recently, the concept of selfsimilar
cascades leading to multifractal measures has been generalized
to the construction of scaleinvariant signals using orthonormal
wavelet basis1922. Instead of reditributing the measure over
subintervals with multiplicative weights, one allocates the
wavelet coefficients in a multiplicative way on a dyadic grid.
This method allows us to generate multifractal functions from
a given deterministic or probabilistic multiplicative process.

 The main goal of this paper is to emphasize the continuous
wavelet transform as a very efficient tool to get deep insight
into the hierarchical structural complexity of multifractal objects23,24.
Our point is to show that the main information about the underlying
cascading process can actually be extracted from the wavelet
transform skeleton defined by the wavelet transform modulus maxima25
(WTMM). We will discuss two situations which are typical, namely
the case of deterministic discrete cascades and the case of random
continuous cascades. For both cases, we will illustrate our purpose
with some applications to experimental data ranging from fractal
growth phenomena to fully developed turbulence and to financial
timeseries.

 Deterministic discrete cascades: In many cases, the selfsimilarity
properties of fractal objects can be expressed in terms of a
dynamical system which leaves the object invariant. The inverse
problem consists in recovering this dynamical system (or its
main characteristics) from the data representing the fractal
object. This problem has been previously approached within the
theory of Iterated Function Systems26 (IFS). But the methods
developed in this context are based on the search for a "bestfit"
within a prescribed class of IFS attractors (mainly linear homogeneous
attractors). In that sense, they approximate the selfsimilarity
properties more than they reveal them. We show that, in many
situations, the WTMM skeleton can be used to extract some onedimensional
(1D) map which accounts for its construction process27,28. We
illustrate our theoretical considerations on pedagogical examples
including Bernoulli invariant measures of linear and nonlinear
expanding Markov maps as well as the invariant measure of perioddoubling
dynamical systems at the onset of chaos. We apply this wavelet
based technique to analyze the fractal properties of DLA azimuthal
Cantor sets defined by intersecting the inner frozen region of
large mass off\lattice Diffusion\Limited\Aggregates (DLA)
with a circle2830. This study clearly reveals the existence
of an underlying multiplicative process that is likely to account
for the Fibonacci structural ordering recently discovered in
the apparently disordered arborescent DLA morphology31. These
results demonstrate the statistical relevance of the golden mean
arithmetic to the fractal hierarchy of DLA azimuthal Cantor sets.

 Random continuous cascades: We start introducing various
statistical quantities such as i) the statistical singularity
spectrum which quantifies the statistical contribution of each
singularity in the signals22; ii) the selfsimilarity kernel14,20,21,32,33
which, from a statistical point of view, characterizes the nature
of the cascade process from a given scale to a finer scale and
iii) the spacescale correlation functions34 which can be proved
to follow a powerlaw behavior when varying the spatial distance
of the two wavelet coefficients. We show mathematically and check
numerically on various computer synthetized signals22,35 that
these statistical quantities can be extracted directly from the
considered fractal function using its WTMM skeleton with an arbitrary
analyzing wavelet. This mathematical study actually provides
algorithms that are readily applicable to experimental situations.
We report on recent applications of our methodology in the context
of fullydeveloped turbulence20,21,34,35. Our results show that
the commonly accepted multifractal description of the intermittency
phenomenon is not valid for finite Reynolds number flows. In
particular, the search for a cascade process reveals a lack of
scaleinvariance for both the velocity and the dissipation fields.
The statistical study of the WTMM yields a very convincing lognormal
law on a well defined range of scales that can be further used
as an objective definition of the inertial range. But in contrast
to the KolmogorovObukhov lognormal cascade model8,9, we find
that the number of cascade steps does not evolve logarithmically
as a function of the scale which is the signature of the breaking
of scaleinvariance20,21. We further comment on the possible
asymptotic validity of the lognormal multifractal description.

 These very first applications are very promising as fas
as further experimental investigations of multilplicative cascade
processes are concerned. It is likely that similar waveletbased
statistical analysis will lead to significant progress in other
fields than fractal growth phenomena and fully developed turbulence.
To conclude, we report on preliminary results of a similar investigation
of financial time series36. Underlying the fluctuations of volatility
(standard deviation) of price variations, we have discovered
the existence of a causal information cascade from large to small
time scales that can be visualized with the wavelet representation.
Let us emphasize that the fact the variations of prices over
one month scale influence in the future the daily prices variations,
is likely to be extraordinarily rich in consequences and this,
not only for the fundamental understanding of the nature of financial
markets, but also (and may be more important) for practical applications.
Indeed, the nature of the correlations across scales that are
implied by this causal cascade have profound implications on
the market risk, a problem of upmost concern for all financial
institutions as well as individuals.

 References

 [1] B.B. Mandelbrot, Fractals: Form, Chance and Dimension
Freeman, San Francisco, 1977); The Fractal Geometry of Nature
(Freeman, San Francisco, 1982).

 [2] G. Paladin et A. Vulpiani, Phys. Rep. 156 (1987) 148.

 [3] C. Meneveau and K.R. Sreenivasan, J. Fluid Mech. 224
(1991) 429.

 [4] H.G.E. Hentschel, Phys. Rev. E 50 (1994) 243.

 [5] U. Frisch, Turbulence (Cambridge Univ. Press, Cambridge,
1995).

 [6] B.B. Mandelbrot, J. Fluid Mech. 62 (1974) 331.

 [7] A.N. Kolmogorov, C.R. Acad. Sci. USSR 30 (1941) 301.

 [8] A.N. Kolmogorov, J. Fluid Mech. 13 (1962) 82.

 [9] A.M. Obukhov, J. Fluid Mech. 13 (1962) 77.

 [10] D. Schertzer, S. Lovejoy, F. Schmitt, Y. Chigirinskaya
and D. Marsan, Fractals 5 (1997) 427.

 [11] E.A. Novikov, Phys. Fluids A 2 (1990) 814; Phys.
Rev. E 50 (1995) 3303.

 [12] Z.S. She and E.C. Waymire, Phys. Rev. Lett. 74 (1995)
262.

 [13] B. Dubrulle, Phys. Rev. Lett. 73 (1994) 959; J. Phys.
II France 6 (1996) 1825.

 [14] B. Castaing and B. Dubrulle, J. Phys. II France 5
(1995) 895.

 [15] Z.S. She and E. Leveque, Phys. Rev. Lett. 72 (1994)
336.

 [16] D. Sornette, in Scale Invariance and Beyond, B. Dubrulle,
F. Graner and D.Sornette, eds (EDP Sciences, Les Ulis & SpringerVerlag,
Berlin, 1997), pp. 235.

 [17] J.P. Kahane and J. Peyrière, Advances in Mathematics
22 (1976) 131.

 [18] G.M. Molchan, Comm. Math. Phys. 179 (1996) 681.

 [19] R. Benzi, L. Biferale, A. Crisanti, G. Paladin, M.
Vergassola and A. Vulpiani, Physica D 65 (1993) 352.

 [20] A. Arneodo, J.F. Muzy and S.G. Roux, J. Phys II France
7 (1997) 363.

 [21] A. Arneodo, S. Manneville and J.F. Muzy, Eur. Phys.
J. B 1 (1998) 129.

 [22] A. Arneodo, E. Bacry and J.F. Muzy, "Random
cascades on wavelet dyadic trees", J. Math. Phys. (1998),
to appear.

 [23] J.F. Muzy, E. Bacry and A. Arneodo, Int. J. of Bifurcation
and Chaos 4 (1994) 245.

 [24] A. Arneodo, E. Bacry and J.F. Muzy, Physica A 213
(1994) 232.

 [25] S. Mallat, A Wavelet Tour in Signal Processing (Academic
Press, New York, 1998).

 [26] M.F. Barnsley, Fractals Everywhere (Academic Press,
New York, 1988).

 [27] A. Arneodo, E. Bacry and J.F. Muzy, Europhys. Lett.
25 (1994) 479.

 [28] A. Arneodo, F. Argoul, J.F. Muzy, M. Tabard and E.
Bacry, Fractals 1 (1993) 629.

 [29] A. Arneodo, F. Argoul, E. Bacry, J.F. Muzy and M.
Tabard, Phys. Rev. Lett. 68 (1992) 3456; J. Diff. Eq. & Appl.
1 (1995) 117.

 [30] A. Kuhn, F. Argoul, J.F. Muzy and A. Arneodo, Phys.
Rev. Lett. 73 (1994) 2998.

 [31] A. Arneodo, F. Argoul, J.F. Muzy and M. Tabard, Phys.
Lett. 171 A (1992) 31; Physica 188 A (1992) 217.

 [32] B. Castaing, Y. Gagne and E.J. Hopfinger, Physica
D 46 (1990) 177.

 [33] B. Castaing, in Scale Invariance and Beyond, B. Dubrulle,
F. Graner and D. Sornette, eds (EDP Sciences, Les Ulis &
SpringerVerlag, Berlin 1997), pp. 225.

 [34] A. Arneodo, E. Bacry, S. Manneville and J.F. Muzy,
Phys. Rev. Lett. 80 (1998) 708.

 [35] S.G. Roux, Thesis, University of AixMarseille II
(1996).

 [36] A. Arneodo, J.F. Muzy and D. Sornette, Eur. Phys.
J. B 1 (1998), to appear.
