12.12.2012 Views

Monte Carlo seismic volcanic envelopes: in need of boundary ...

Monte Carlo seismic volcanic envelopes: in need of boundary ...

Monte Carlo seismic volcanic envelopes: in need of boundary ...

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

submitted to Geophys. J. Int.<br />

<strong>Monte</strong> <strong>Carlo</strong> <strong>seismic</strong> <strong>volcanic</strong> <strong>envelopes</strong>: <strong>in</strong> <strong>need</strong> <strong>of</strong> <strong>boundary</strong><br />

conditions<br />

L. De Siena 1 E. Del Pezzo 2,3 C. Thomas 1 A. Curtis 4 L. Marger<strong>in</strong> 5<br />

1 University <strong>of</strong> Münster, Institute for Geophysics, Correnstrasse 24,<br />

Münster, 48149, Germany. (lucadesiena@uni-muenster.de)<br />

2 Istituto Nazionale di Ge<strong>of</strong>isica e Vulcanologia, Department <strong>of</strong> Naples -<br />

Osservatorio Vesuviano, Via Diocleziano 328, Naples, 80124, Italy.<br />

3 also at Istituto Andaluz de Ge<strong>of</strong>isica, Universidad de Granada, Granada, Spa<strong>in</strong>.<br />

4 School <strong>of</strong> Geosciences, University <strong>of</strong> Ed<strong>in</strong>burgh, Ed<strong>in</strong>burgh<br />

Scotland.<br />

5 Ludovic Marger<strong>in</strong>. Institut de Recherche en Astrophysique et<br />

Plantologie, Université Paul Sabatier / CNRS, 14 Avenue Edouard<br />

Bel<strong>in</strong>, 31400 Toulouse, France.<br />

January 20, 2012<br />

SUMMARY<br />

A determ<strong>in</strong>istic and/or a stochastic technique can be pr<strong>of</strong>iciently applied on seismological data<br />

if the limit between coherent and <strong>in</strong>coherent signals is clearly marked both <strong>in</strong> time and <strong>in</strong> fre-<br />

quency. We can exploit the signal to provide adequate images <strong>of</strong> the Earth only if we know its<br />

degree <strong>of</strong> coherency. Nevertheless, <strong>in</strong> <strong>volcanic</strong> <strong>seismic</strong> record<strong>in</strong>gs, the mix<strong>in</strong>g between coher-<br />

ent and <strong>in</strong>coherent signals is cont<strong>in</strong>uous. We account for the coherence effects <strong>in</strong> a <strong>volcanic</strong><br />

<strong>seismic</strong> envelope <strong>in</strong>clud<strong>in</strong>g <strong>in</strong> the usual stochastic approach a drastic change <strong>in</strong> the scatterer<br />

distribution. Even if caused by a determ<strong>in</strong>istic <strong>boundary</strong>, the effects <strong>of</strong> this change are still<br />

stochastic <strong>in</strong> their mathematical and physical form, and can be described by a second set <strong>of</strong><br />

transport equations. Coda signals are triggered <strong>in</strong> our model by three different length scales:


1<br />

2<br />

3<br />

4<br />

5<br />

6<br />

7<br />

8<br />

9<br />

10<br />

11<br />

2 L. De Siena E. Del Pezzo C. Thomas A. Curtis L. Marger<strong>in</strong><br />

the mean free path, the transport mean free path, and the distance between the source and a<br />

large scale <strong>boundary</strong>. Above 10 Hz, we can model our <strong>in</strong>tensity on real <strong>envelopes</strong> recorded<br />

at Campi Flegrei caldera without the <strong>boundary</strong>. This as well as the reach <strong>of</strong> diffusion after<br />

three times the S-wave arrival corroborates the use <strong>of</strong> the ray approximation and <strong>of</strong> the coda<br />

normalization method, respectively. Even <strong>in</strong> this high frequency limit, unexpected large SP<br />

conversions and general broaden<strong>in</strong>g are evident <strong>in</strong> the <strong>in</strong>termediate coda. Around 1 Hz, the<br />

unexpected broaden<strong>in</strong>g affects the whole envelope. Diffraction and <strong>in</strong>terference effects are rel-<br />

evant, while recursive coherent signals arise at <strong>in</strong>termediate times. We implement a <strong>boundary</strong><br />

<strong>in</strong> the simulation through the def<strong>in</strong>ition <strong>of</strong> a different scatter<strong>in</strong>g texture, co<strong>in</strong>cident with the<br />

caldera rim signature at 1.5 km. The <strong>boundary</strong> acts as a filter between coherent and <strong>in</strong>coherent<br />

<strong>in</strong>tensities, so that 2D synthetics provide an adequate first order model <strong>of</strong> the data <strong>envelopes</strong><br />

also <strong>in</strong> the low frequency range. A model based on the theory <strong>of</strong> <strong>in</strong>homogeneous dispersive<br />

turbulence should be implemented to model low frequency <strong>envelopes</strong> <strong>in</strong> the whole caldera. We<br />

proved, anyway, that the exclusive application <strong>of</strong> a determ<strong>in</strong>istic or stochastic approximation<br />

provides unfeasible results, while a mixed approach can better describe the envelope behaviors<br />

<strong>in</strong> the whole frequency range.<br />

Key words: Radiative Transfer Theory – Large angle scatter<strong>in</strong>g – <strong>Monte</strong> <strong>Carlo</strong> simulation<br />

1 INTRODUCTION<br />

Radiative transfer theory describes the transport <strong>of</strong> <strong>seismic</strong> energy through a scatter<strong>in</strong>g medium; it<br />

is one <strong>of</strong> the most important tools, together with diffusion theory, to synthesize coda <strong>in</strong>tensities, the<br />

ma<strong>in</strong> evidence <strong>of</strong> scatter<strong>in</strong>g <strong>in</strong> the Earth (Wu, 1985). Radiative transfer solutions provide <strong>in</strong>tensity<br />

variations comparable with coda <strong>envelopes</strong>; these <strong>in</strong>tensity measurements, usually calculated as<br />

the mean square or the root mean square <strong>of</strong> coda signals, are crucial measurements <strong>of</strong> the degree<br />

<strong>of</strong> <strong>in</strong>homogeneity <strong>of</strong> the Earth (Sato & Fehler, 1998). For scalar waves, radiative transfer theory<br />

only provides two exact analytical solutions: the stationary and the 2-D dynamic solutions for<br />

isotropic scatter<strong>in</strong>g. Otherwise, radiative transfer equations are solved numerically (i. e. with a<br />

<strong>Monte</strong> <strong>Carlo</strong> approach) (Ishimaru, 1997).<br />

A rigorous multiple scatter<strong>in</strong>g analytical theory has been developed <strong>in</strong> the past 20 years to de-


12<br />

13<br />

14<br />

15<br />

16<br />

17<br />

18<br />

19<br />

20<br />

21<br />

22<br />

23<br />

24<br />

25<br />

26<br />

27<br />

28<br />

29<br />

30<br />

31<br />

32<br />

33<br />

34<br />

35<br />

36<br />

37<br />

38<br />

39<br />

<strong>Monte</strong> <strong>Carlo</strong> <strong>seismic</strong> <strong>envelopes</strong> caldera 3<br />

scribe <strong>seismic</strong> <strong>in</strong>tensities (Zeng et al., 1991; Zeng, 1991, 1993; Sato & Fehler, 1998). In practice,<br />

this theory only provides approximate solutions, particularly for complex <strong>volcanic</strong> media. Hence,<br />

s<strong>in</strong>gle scatter<strong>in</strong>g and diffusion are used to measure critical scatter<strong>in</strong>g parameters, such as the trans-<br />

port mean free path, or the ratio between mean square fluctuations and correlation distance (Sato &<br />

Fehler, 1998; Wegler & Lühr, 2001; Przybilla et al., 2009). S<strong>in</strong>gle scatter<strong>in</strong>g and diffusion are limit<br />

cases <strong>of</strong> multiple scatter<strong>in</strong>g. When deal<strong>in</strong>g with a tenuous distribution <strong>of</strong> particles, s<strong>in</strong>gle scatter-<br />

<strong>in</strong>g isotropic solutions are used to model energy <strong>envelopes</strong> at regional scale (Wu, 1985; Przybilla<br />

et al., 2006). The s<strong>in</strong>gle scatter<strong>in</strong>g approximation gives <strong>in</strong>correct results <strong>in</strong> a medium with strong<br />

velocity fluctuations (Sato & Fehler, 1998); diffusion theory is therefore a more suitable candidate<br />

to describe scatter<strong>in</strong>g propagation <strong>in</strong> <strong>volcanic</strong> areas (Wegler & Lühr, 2001). Nevertheless, for short<br />

source-receiver distances (i. e. <strong>in</strong> volcano monitor<strong>in</strong>g) diffusion could be an over-estimation <strong>of</strong> the<br />

<strong>in</strong>coherence characteriz<strong>in</strong>g the <strong>seismic</strong> wave-field (Yamamoto & Sato, 2010).<br />

The waves scattered by random heterogeneities are never totally <strong>in</strong>coherent; the coherent field<br />

goes to zero exponentially with time, but is still part <strong>of</strong> the diffusion solution (Ishimaru, 1997).<br />

The underly<strong>in</strong>g <strong>in</strong>tr<strong>in</strong>sic coherence <strong>in</strong> <strong>volcanic</strong> signals is due to the presence <strong>of</strong> strong velocity<br />

fluctuations caus<strong>in</strong>g multiple scatter<strong>in</strong>g, that become important at high frequencies and at short<br />

lapse-times (Sato, 1989). Phenomenologically, these fluctuations broaden <strong>in</strong> time and lower <strong>in</strong><br />

amplitude coda <strong>envelopes</strong>; both the s<strong>in</strong>gle scatter<strong>in</strong>g model and the diffusion theory cannot expla<strong>in</strong><br />

these phenomena (Sato & Fehler, 1998; Saito et al., 2005).<br />

The Markov approximation for the parabolic wave equation is the usual approach to account<br />

for broaden<strong>in</strong>g and maximum amplitude decay (Sato, 1989; Fehler et al., 2000; Saito et al., 2002).<br />

The parabolic wave equation models the scatter<strong>in</strong>g contribution <strong>of</strong> the long-wavelength spectra. Its<br />

Markov approximation is valid <strong>in</strong> case <strong>of</strong> strong forward scatter<strong>in</strong>g both for plane and for spher-<br />

ical waves. Markov solutions - also known as Markov <strong>envelopes</strong> - require the <strong>in</strong>troduction <strong>of</strong> a<br />

two-frequency mutual coherence function, represent<strong>in</strong>g the background correlation <strong>of</strong> the velocity<br />

field (Sato & Fehler, 1998; Fehler et al., 2000; Saito et al., 2002). The Markov approximation<br />

is used directly or <strong>in</strong> a f<strong>in</strong>ite frequency simulation to synthesize Markov <strong>envelopes</strong> (Sato et al.,<br />

2004; Saito et al., 2005). Sato et al. (2004) particularly remark that s<strong>in</strong>gle scatter<strong>in</strong>g coefficients


40<br />

41<br />

42<br />

43<br />

44<br />

45<br />

46<br />

47<br />

48<br />

49<br />

50<br />

51<br />

52<br />

53<br />

54<br />

55<br />

56<br />

57<br />

58<br />

59<br />

60<br />

61<br />

62<br />

63<br />

64<br />

65<br />

66<br />

67<br />

4 L. De Siena E. Del Pezzo C. Thomas A. Curtis L. Marger<strong>in</strong><br />

cannot efficiently reproduce broaden<strong>in</strong>g, s<strong>in</strong>ce they produce large forward scatter<strong>in</strong>g. Hence, these<br />

authors <strong>in</strong>troduce the momentum scatter<strong>in</strong>g coefficients to describe statistically a medium where<br />

they propagate the Markov solution. Nevertheless, <strong>seismic</strong> <strong>envelopes</strong> cannot be entirely repro-<br />

duced without superimpos<strong>in</strong>g large-angle scatter<strong>in</strong>g at small-scale random heterogeneities, as well<br />

as short-wavelength scatter<strong>in</strong>g (Fukushima et al., 2003; Sato et al., 2004; Przybilla et al., 2006).<br />

Except when analytic solutions exist, <strong>Monte</strong> <strong>Carlo</strong> numerical simulations <strong>of</strong> the radiative trans-<br />

fer equations are used to synthesize <strong>seismic</strong> <strong>envelopes</strong> (Gusev & Abubakirov, 1987; Hoshiba,<br />

1991; Marger<strong>in</strong> et al., 1998; Yoshimoto, 2000). Polarization <strong>in</strong>formation and broaden<strong>in</strong>g <strong>of</strong> the<br />

envelope are <strong>in</strong>cluded <strong>in</strong> the techniques <strong>in</strong> order to obta<strong>in</strong> solutions modeled on real <strong>envelopes</strong><br />

(Marger<strong>in</strong> et al., 2000; Przybilla et al., 2006; Przybilla & Korn, 2008). At first aimed at the de-<br />

scription <strong>of</strong> energy propagation at regional scale, these techniques also provide important results<br />

at tele<strong>seismic</strong> distances, e. g. through the analysis <strong>of</strong> PKP precursors (Marger<strong>in</strong> & Nolet, 2003).<br />

Local scale simulations present different challenges. Marger<strong>in</strong> & van Tiggelen (2001) and van<br />

Tiggelen et al. (2001) study the coda <strong>in</strong>tensity enhancement due to <strong>in</strong>terference <strong>in</strong> the near field.<br />

They highlight a reciprocal wave produc<strong>in</strong>g constructive <strong>in</strong>terference (coherence) after a transient<br />

regime, likewise a stable spherical source <strong>of</strong> radius half a wavelength, centered at the <strong>seismic</strong><br />

source (Marger<strong>in</strong> & van Tiggelen, 2001). The theory is valid if the source-receiver distance is<br />

with<strong>in</strong> approximately one elastic wavelength, as can be the case for a monitored <strong>volcanic</strong> area, but<br />

is unlikely to be observed for earthquake sources due to the broken reciprocity between source and<br />

receiver (van Tiggelen et al., 2001). It is evident that at each scale, and for different waves, <strong>Monte</strong><br />

<strong>Carlo</strong> solutions improve the understand<strong>in</strong>g <strong>of</strong> the physics underlay<strong>in</strong>g scatter<strong>in</strong>g propagation.<br />

We implement a full 2D elastic <strong>Monte</strong> <strong>Carlo</strong> simulation <strong>of</strong> the <strong>in</strong>tensity produced by a po<strong>in</strong>t<br />

source <strong>in</strong> a <strong>volcanic</strong> medium through a priori def<strong>in</strong>ition <strong>of</strong> the s<strong>in</strong>gle and momentum scatter-<br />

<strong>in</strong>g coefficients, us<strong>in</strong>g a von Kármán autocorrelation function (Sato et al., 2004; Przybilla et al.,<br />

2006). The coefficients trigger both the probability and the propagation direction after each scat-<br />

ter<strong>in</strong>g event. Direct-wave energy reproduction, reach <strong>of</strong> diffusion, l<strong>in</strong>e-<strong>of</strong>- sight propagation, and<br />

early-coda coherency are <strong>in</strong>vestigated compar<strong>in</strong>g the synthetics with the real <strong>envelopes</strong> recoded<br />

at Campi Flegrei caldera. A large change <strong>in</strong> the distribution <strong>of</strong> the scatterers is <strong>in</strong>cluded to model


68<br />

69<br />

70<br />

71<br />

72<br />

73<br />

74<br />

75<br />

76<br />

77<br />

78<br />

79<br />

80<br />

81<br />

82<br />

83<br />

84<br />

85<br />

86<br />

87<br />

88<br />

89<br />

90<br />

91<br />

<strong>Monte</strong> <strong>Carlo</strong> <strong>seismic</strong> <strong>envelopes</strong> caldera 5<br />

our solutions on real coda <strong>envelopes</strong>. We <strong>in</strong>clude the enhancement <strong>of</strong> large angle scatter<strong>in</strong>g by the<br />

def<strong>in</strong>ition <strong>of</strong> a reverse scatter<strong>in</strong>g wave-field (reverse respect to the <strong>in</strong>cident direction) created by<br />

the <strong>boundary</strong>. Our scope is to obta<strong>in</strong> a first order model <strong>of</strong> the real <strong>envelopes</strong> <strong>in</strong> order to depict the<br />

mechanism produc<strong>in</strong>g broaden<strong>in</strong>g <strong>in</strong> <strong>volcanic</strong> areas (Yamamoto & Sato, 2010).<br />

2 MONTE CARLO SIMULATION OF MULTIPLE SCATTERING PROPAGATION<br />

Multiple scatter<strong>in</strong>g theory describes wave propagation <strong>in</strong> a tenuous distribution <strong>of</strong> scatterers via a<br />

s<strong>in</strong>gle scatter<strong>in</strong>g approximation. The s<strong>in</strong>gle scatter<strong>in</strong>g coefficients provide the scatter<strong>in</strong>g pattern at<br />

each scatter<strong>in</strong>g event; multiple scatter<strong>in</strong>g then produces the entire coda envelope. This approach<br />

was applied to synthesize <strong>envelopes</strong> recorded at long lapse times (late coda) for regional distances<br />

(Sato et al., 2004; Przybilla et al., 2006). Hence, s<strong>in</strong>gle scatter<strong>in</strong>g is the first step for understand<strong>in</strong>g<br />

multiple scatter<strong>in</strong>g processes.<br />

We aim to ga<strong>in</strong> an <strong>in</strong>sight <strong>in</strong>to the decay with time <strong>of</strong> <strong>envelopes</strong> recorded at local scale and for<br />

<strong>volcanic</strong> regions. Therefore, we solve the radiative transfer equations by the <strong>Monte</strong> <strong>Carlo</strong> method<br />

<strong>in</strong> the full elastic case, us<strong>in</strong>g a s<strong>in</strong>gle scatter<strong>in</strong>g approximation (Marger<strong>in</strong> et al., 2000; Przybilla<br />

et al., 2006). This is just a basement stone; the small source-receiver distances and the high hetero-<br />

geneity <strong>of</strong> the velocity wave-field <strong>in</strong> a <strong>volcanic</strong> area can break the s<strong>in</strong>gle scatter<strong>in</strong>g approximation,<br />

only plausible at larger scales (Przybilla et al., 2006).<br />

2.1 Statistical description <strong>of</strong> the medium<br />

A statistical description <strong>of</strong> the medium through its correlation and spectral aspects is necessary<br />

to describe scatter<strong>in</strong>g propagation (Rytov et al., 1987). We consider a crustal area <strong>of</strong> average<br />

dimension L = 10 km, which can be described as an ensemble <strong>of</strong> random media, and characterized<br />

by a 2D random velocity field (Figure 1). This field V (x) is described <strong>in</strong> terms <strong>of</strong> average wave<br />

velocity V0 and the fractional velocity fluctuation ξ(x), where x is the 2D space coord<strong>in</strong>ate (Sato<br />

& Fehler, 1998):<br />

V (x) = V0{1 + ξ(x)}. (1)


92<br />

93<br />

94<br />

95<br />

96<br />

97<br />

98<br />

99<br />

100<br />

101<br />

102<br />

103<br />

104<br />

105<br />

106<br />

107<br />

108<br />

109<br />

110<br />

111<br />

112<br />

113<br />

6 L. De Siena E. Del Pezzo C. Thomas A. Curtis L. Marger<strong>in</strong><br />

We def<strong>in</strong>e R(x), the auto-correlation function (ACF) <strong>of</strong> the fractional velocity fluctuations between<br />

po<strong>in</strong>ts <strong>of</strong> positions x and x’, as:<br />

R(x) ≡ 〈ξ(x+x’)ξ(x’)〉, (2)<br />

and its Fourier transform, the power spectral density function (PSDF), as:<br />

P (k) =<br />

�∞<br />

−∞<br />

R(x) exp(−ikx)dx, (3)<br />

where k is the wave number, and angle brackets stand for ensemble average. The PSDF is the key<br />

function to describe scatter<strong>in</strong>g; it must be carefully chosen <strong>in</strong> order to represent real <strong>envelopes</strong>.<br />

A power low PSDF corresponds to an exponential or, more generally, a von Kármán-type ACF<br />

(Shapiro & Hubral, 1999). The von Kármán PSDF is:<br />

P (k) =<br />

4πɛ2a2κ (1 + a2k 2 , (4)<br />

) (κ+1)<br />

where a is the correlation length, ɛ 2 ≡ R(0) = 〈ξ(x) 2 〉 is the mean square fluctuation, Γ is the<br />

Gamma function, and κ is the order <strong>of</strong> the function, which triggers the decay <strong>of</strong> the PSDF. These<br />

quantities commonly characterize a random distribution <strong>of</strong> particles; <strong>in</strong> seismology, an exponential<br />

PSDF is preferable to a Gaussian one <strong>in</strong> order to depict <strong>seismic</strong> <strong>envelopes</strong>. This is particularly<br />

true at local scale, as suggested by well-log and <strong>seismic</strong> data (Shiomi et al., 1997). S<strong>in</strong>ce we<br />

deal with a <strong>volcanic</strong> medium, we assume that long-wavelength spectra are dom<strong>in</strong>at<strong>in</strong>g, caus<strong>in</strong>g<br />

broaden<strong>in</strong>g, loss <strong>of</strong> coherence, and decrease <strong>in</strong> mean field <strong>in</strong>tensity. Multiple scatter<strong>in</strong>g simulations<br />

with s<strong>in</strong>gle scatter<strong>in</strong>g coefficients and with an exponential (κ = 0.5) ACF provide <strong>envelopes</strong> at<br />

regional distance, witch are similar to the real one. To account for the statistical long wavelength<br />

effects we also consider the highest order <strong>of</strong> the ACF (κ = 1). The PSDFs for both orders are<br />

shown <strong>in</strong> Figure 2.<br />

2.2 S<strong>in</strong>gle scatter<strong>in</strong>g coefficients<br />

In seismology, 2D and 3D <strong>Monte</strong> <strong>Carlo</strong> solutions reproduce the <strong>envelopes</strong> <strong>of</strong> real <strong>seismic</strong> P- and S-<br />

waves propagat<strong>in</strong>g through the Lithosphere both <strong>in</strong> the acoustic and <strong>in</strong> the elastic cases (Gusev &<br />

Abubakirov, 1996; Marger<strong>in</strong> et al., 2000; Przybilla et al., 2009). We simulate the multiply scattered


114<br />

115<br />

116<br />

117<br />

118<br />

119<br />

120<br />

121<br />

122<br />

123<br />

124<br />

125<br />

126<br />

127<br />

128<br />

129<br />

130<br />

131<br />

<strong>Monte</strong> <strong>Carlo</strong> <strong>seismic</strong> <strong>envelopes</strong> caldera 7<br />

wave-field by shoot<strong>in</strong>g particles from a po<strong>in</strong>t source located <strong>in</strong> the caldera, whose velocity field is<br />

statistically characterized by a von Kármán PSDF. Collision <strong>of</strong> the particle with the scatterers are<br />

perfectly elastic; the scattered direction, chosen with a table-look up method, is anisotropic, and is<br />

triggered by the differential s<strong>in</strong>gle scatter<strong>in</strong>g coefficients, equivalent to the differential scatter<strong>in</strong>g<br />

cross-sections (Przybilla et al., 2006).<br />

The s<strong>in</strong>gle differential scatter<strong>in</strong>g coefficients are dependent on the von Kármán PSDF as well<br />

as on the scatter<strong>in</strong>g patterns, |Xij(θ) 2 |; here θ is the angle respect to the <strong>in</strong>cident direction, while<br />

ij stands for P or S (Sato & Fehler, 1998). The SS differential scatter<strong>in</strong>g coefficient gss is given<br />

by:<br />

gss(θ) = k3 s<br />

8π P (2ks s<strong>in</strong>( θ<br />

2 ))|Xss(θ)| 2 , (5)<br />

where ks is the S-wavenumber, density is correlated to velocity by the factor ν = 0.8 (given by<br />

Birch’s law), and Xss(θ) is def<strong>in</strong>ed by:<br />

Xss(θ) = (ν(cos θ − cos 2θ) − 2 cos 2θ). (6)<br />

SS scatter<strong>in</strong>g dom<strong>in</strong>ates for large lapse-times <strong>in</strong> the cont<strong>in</strong>ental Lithosphere - for a complete review<br />

on the variations <strong>of</strong> the scatter<strong>in</strong>g coefficients with space, time, and frequency see Sato & Fehler<br />

(1998). In this paper, we refer to their 2D equivalent, extensively treated by Przybilla et al. (2006).<br />

P- and S-wave mean free paths (lp and ls, respectively) are the average distances traveled by a<br />

particle between two collisions. They are dependent on the s<strong>in</strong>gle scatter<strong>in</strong>g coefficients:<br />

We <strong>in</strong>troduce the notations:<br />

g 0 ij = 1<br />

2π<br />

� 2π<br />

0<br />

g 0 p = g 0 pp + g 0 ps<br />

where the cross-terms satisfies the 2-D reciprocity relation:<br />

gij(θ)dθ. (7)<br />

(8a)<br />

g 0 s = g 0 sp + g 0 ss, (8b)<br />

g 0 ps = ( Vp<br />

Vs<br />

(8c)<br />

) 3<br />

2 g 0 sp, (9)


132<br />

133<br />

134<br />

135<br />

136<br />

137<br />

138<br />

139<br />

140<br />

141<br />

142<br />

143<br />

144<br />

145<br />

146<br />

147<br />

148<br />

149<br />

150<br />

151<br />

152<br />

153<br />

154<br />

155<br />

8 L. De Siena E. Del Pezzo C. Thomas A. Curtis L. Marger<strong>in</strong><br />

and Vp and Vs are P- and S-wave velocities; hence, the mean free paths are:<br />

lp = (g 0 p) −1<br />

(10a)<br />

ls = (g 0 s) −1 ; (10b)<br />

The scatter<strong>in</strong>g coefficients depend on the typical correlation distance for a <strong>volcanic</strong> area (a);<br />

ACF velocity measurements at Campi Flegrei provide a value <strong>of</strong> a = 0.9 (De Siena et al., 2011).<br />

We assume a mean squared fluctuation ɛ = 0.06%, but a smaller correlation length (a ∼ = 0.5) <strong>in</strong><br />

order to account for the highly heterogeneous S-wave velocity field <strong>in</strong> the caldera (Battaglia et al.,<br />

2008). This assumption is supported by K. & Levander (1992), who found an average correlation<br />

length for the cont<strong>in</strong>ental crust between 0.2 km and 0.8 km. We also <strong>need</strong> to lower a to model<br />

both P- and S-wave propagation by us<strong>in</strong>g geometrical optics; a larger correlation length leads to a<br />

lS < a, and the wave would be attenuated before go<strong>in</strong>g through a s<strong>in</strong>gle heterogeneity.<br />

The s<strong>in</strong>gle scatter<strong>in</strong>g coefficients numerically calculated for 3 Hz and 18 Hz are shown <strong>in</strong><br />

Table 1; a low probability for conversion scatter<strong>in</strong>g still exists even at high frequency. We numer-<br />

ically compute the scatter<strong>in</strong>g pattern associated with each scatter<strong>in</strong>g mode to select the direction<br />

after each scatter<strong>in</strong>g event (θ <strong>in</strong> Figure 1) with a table-look up method. We obta<strong>in</strong> different tables<br />

<strong>of</strong> angles for different scatter<strong>in</strong>g modes (Table 2). Each <strong>of</strong> the 201 angles corresponds to an equal<br />

fraction <strong>of</strong> the total cross section, given by the <strong>in</strong>tegral over 2π <strong>of</strong> the scatter<strong>in</strong>g coefficients (Equa-<br />

tion (5)). Table 2 is an example <strong>of</strong> the angle distribution for s<strong>in</strong>gle SS and PS scatter<strong>in</strong>g, κ = 0.5,<br />

and frequency 3 Hz. Direction is selected by us<strong>in</strong>g a random <strong>in</strong>teger number (def<strong>in</strong>ed between<br />

1 and 201): it is evident that scatter<strong>in</strong>g is most likely <strong>in</strong> the forward direction for SS scatter<strong>in</strong>g,<br />

while the PS (and SP) scatter<strong>in</strong>g patterns have their ma<strong>in</strong> lobes <strong>in</strong> transversal directions. Conver-<br />

sion scatter<strong>in</strong>g <strong>in</strong>duces a relevant deviation from the <strong>in</strong>cident direction, caus<strong>in</strong>g a larger sampl<strong>in</strong>g<br />

<strong>of</strong> the medium (Figure 1).<br />

2.3 Momentum scatter<strong>in</strong>g coefficients and transport mean free path<br />

In <strong>volcanic</strong> regions, multiple scatter<strong>in</strong>g at large lapse times is considered as isotropic, even if each<br />

scatter<strong>in</strong>g process is not (Yamamoto & Sato, 2010). Some authors prefer to simulate <strong>envelopes</strong>


156<br />

157<br />

158<br />

159<br />

160<br />

161<br />

162<br />

163<br />

164<br />

165<br />

166<br />

167<br />

168<br />

169<br />

170<br />

171<br />

172<br />

173<br />

174<br />

175<br />

176<br />

177<br />

<strong>Monte</strong> <strong>Carlo</strong> <strong>seismic</strong> <strong>envelopes</strong> caldera 9<br />

at large lapse times exclud<strong>in</strong>g strong forward scatter<strong>in</strong>g from the anisotropic s<strong>in</strong>gle scatter<strong>in</strong>g<br />

coefficients. This can be done <strong>in</strong>clud<strong>in</strong>g the factor (1 − cos θ) <strong>in</strong> the def<strong>in</strong>ition <strong>of</strong> gij (Gusev &<br />

Abubakirov, 1996; Sato et al., 2004); the SS differential momentum scatter<strong>in</strong>g coefficients (gss m )<br />

are therefore def<strong>in</strong>ed us<strong>in</strong>g Equation (5):<br />

g m ss(θ) = (1 − cos θ) k3 S<br />

8π P (2ks s<strong>in</strong> θ<br />

2 )|Xss(θ)|, (11)<br />

and represent a background medium rich <strong>in</strong> short-wavelength <strong>in</strong>homogeneities (Table 3).<br />

First <strong>in</strong>troduced by Gusev & Abubakirov (1996) <strong>in</strong> seismology for characteriz<strong>in</strong>g coda exci-<br />

tation at long lapse times, they were then applied by Sato et al. (2004) as background medium<br />

for propagat<strong>in</strong>g waves with a diffraction impr<strong>in</strong>t<strong>in</strong>g - named Markov <strong>envelopes</strong>. In <strong>volcanic</strong> areas,<br />

these <strong>envelopes</strong> correlate better than the ones obta<strong>in</strong>ed with s<strong>in</strong>gle scatter<strong>in</strong>g coefficients to the<br />

result <strong>of</strong> the <strong>in</strong>version <strong>of</strong> <strong>seismic</strong> <strong>in</strong>tensity produced by active sources (Yamamoto & Sato, 2010).<br />

Nevertheless, it is known s<strong>in</strong>ce the works <strong>of</strong> Weaver (1990) and Ryzhik et al. (1996) that, <strong>in</strong> the<br />

elastic case, the <strong>in</strong>clusion <strong>of</strong> a factor (1 − cos θ) is not preserv<strong>in</strong>g total <strong>in</strong>tensity. Also, the equiva-<br />

lent scale lengths (the transport mean free paths) have more complicated expressions than the ones<br />

obta<strong>in</strong>ed from momentum scatter<strong>in</strong>g coefficients, at least <strong>in</strong> the full elastic case.<br />

We <strong>in</strong>troduce the cos<strong>in</strong>e weighted scatter<strong>in</strong>g coefficient for the various mode conversions g ∗ ij<br />

(Table 4) as:<br />

g ∗ ij(θ) = 1<br />

� 2π<br />

gij(θ) cos θdθ. (12)<br />

2π 0<br />

Us<strong>in</strong>g these expressions, we obta<strong>in</strong> the transport mean free paths (l ∗ p and l ∗ s, Table 5) <strong>in</strong> the full<br />

elastic case (Turner, 1998):<br />

l ∗ p =<br />

l ∗ s =<br />

g 0 s − g ∗ ss + g ∗ ps<br />

(g 0 p − g ∗ pp)(g 0 s − g ∗ ss) − g ∗ps g ∗ sp<br />

g 0 p − g ∗ pp + g ∗ sp<br />

(13a)<br />

(g0 p − g∗ pp)(g0 s − g∗ ss) − g ∗ps g∗ . (13b)<br />

sp<br />

In the case <strong>of</strong> non-preferential scatter<strong>in</strong>g (i.e. equal amount <strong>of</strong> forward and backward scatter<strong>in</strong>g),<br />

the transport mean free paths reduce to the mean free paths. The transport mean free paths def<strong>in</strong>ed<br />

us<strong>in</strong>g the momentum scatter<strong>in</strong>g coefficients (l m P and lm S<br />

) result larger for S- than for P- waves<br />

(Table 5). This is unfair, s<strong>in</strong>ce we expect a greater level <strong>of</strong> <strong>in</strong>teraction for S-waves, correspond<strong>in</strong>g


178<br />

179<br />

180<br />

181<br />

182<br />

183<br />

184<br />

185<br />

186<br />

187<br />

188<br />

189<br />

190<br />

191<br />

192<br />

193<br />

194<br />

195<br />

196<br />

197<br />

198<br />

199<br />

200<br />

10 L. De Siena E. Del Pezzo C. Thomas A. Curtis L. Marger<strong>in</strong><br />

to a fastest reach <strong>of</strong> diffusion. The opposite happens us<strong>in</strong>g the cos<strong>in</strong>e weighted coefficients. Even<br />

if the diffusion constants (D m and D, respectively) are similar at 18 Hz, they differ <strong>in</strong> the 3 Hz<br />

frequency band. From these considerations as well as for the more sounded connection with the<br />

physics <strong>of</strong> the problem, we will use l ∗ p,s to <strong>in</strong>vestigate diffusion.<br />

2.4 Diffusion<br />

In section 4 we employ the momentum and cos<strong>in</strong>e weighted scatter<strong>in</strong>g coefficients to depict the<br />

reflection on a large-scale <strong>boundary</strong> <strong>in</strong>cluded <strong>in</strong> the medium. However the relative transport mean<br />

free paths are the step length <strong>of</strong> an equivalent diffusion process, hence, a marker <strong>of</strong> the reach<br />

<strong>of</strong> diffusion. Most <strong>of</strong> the techniques employed to measure scatter<strong>in</strong>g parameters and to monitor<br />

<strong>volcanic</strong> area without passive sources only work <strong>in</strong> the diffusion regime (Wegler, 2003; Brenguier<br />

et al., 2008).<br />

Diffusion represents the scatter<strong>in</strong>g radiation field at large lapse times from the source enucle-<br />

ation time. It assumes an energy flux scattered <strong>in</strong> space almost uniformly to every direction, and<br />

recorded after encounter<strong>in</strong>g many particles. A slightly anisotropic angular dependence (like the<br />

one obta<strong>in</strong>ed from momentum or cos<strong>in</strong>e weighted scatter<strong>in</strong>g coefficients) is necessary to add net<br />

power propagation, still present <strong>in</strong> the diffusion regime. Diffuse <strong>in</strong>tensity (ID) at distance r can<br />

therefore be written as the sum <strong>of</strong> the average diffuse <strong>in</strong>tensity (UD(r)) and the diffuse flux vector<br />

(FD(r)), whose direction is given by a unit vector sf:<br />

ID(r, s) ∼ = UD(r) + 3<br />

4π FD(r)sf. (14)<br />

For diffusion to be valid, the second term must be much smaller than the first.<br />

Diffuse energy (ED(r, t)) at distance r <strong>in</strong> the case <strong>of</strong> 2D propagation is given by:<br />

ED(r, t) =<br />

exp(− r2<br />

4Dt )<br />

4πDt<br />

Likewise <strong>in</strong> 3D, 2D diffusion <strong>of</strong> P- and S-waves requires equipartition to be reached (Henn<strong>in</strong>o<br />

et al., 2001; Marger<strong>in</strong>, 2006):<br />

Es<br />

Ep<br />

= V 2<br />

p<br />

V 2<br />

s<br />

(15)<br />

, (16)<br />

where Ep and Es are the energies for P- and S- waves. We rem<strong>in</strong>d that, <strong>in</strong> 2D, P-waves can only


201<br />

202<br />

203<br />

204<br />

205<br />

206<br />

207<br />

208<br />

209<br />

210<br />

211<br />

212<br />

213<br />

214<br />

215<br />

216<br />

217<br />

218<br />

219<br />

220<br />

221<br />

222<br />

223<br />

224<br />

225<br />

<strong>Monte</strong> <strong>Carlo</strong> <strong>seismic</strong> <strong>envelopes</strong> caldera 11<br />

couple with SV-waves. Hence, from previous expressions <strong>of</strong> the diffusivity <strong>of</strong> elastic waves (D) <strong>in</strong><br />

3-D (Weaver, 1990; Ryzhik et al., 1996), we deduce the follow<strong>in</strong>g 2D diffusivity:<br />

D =<br />

1<br />

1 + 2γ2 [Vpl ∗ p<br />

2 + γ2Vsl ∗ s<br />

2<br />

], (17)<br />

where γ = Vp<br />

Vs . The diffusivity (Dm and D), the mean free paths, and the transport mean free<br />

paths are reported <strong>in</strong> Table 5 for different frequencies and κ = 0.5. Figure 3 shows the diffuse<br />

<strong>in</strong>tensities given by Equation 15, at a distance <strong>of</strong> 4 km, and us<strong>in</strong>g the diffusivity D. At both<br />

frequencies, the solutions decay exponentially after a few seconds; they are compared with the<br />

synthetic <strong>in</strong>tensities, obta<strong>in</strong>ed through the <strong>Monte</strong> <strong>Carlo</strong> simulation for a s<strong>in</strong>gle source <strong>in</strong> section<br />

2.5. Before this comparison, anyway, some considerations are necessary<br />

The time soil at which diffusion is reached is dependent on topography (Sanchez-Sesma,<br />

1993). We concentrate anyway on area almost devoid <strong>of</strong> topography; if compared with <strong>volcanic</strong><br />

cones, where topography-effects may <strong>in</strong>duce a faster diffusion reach (Wegler, 2003) or coda-<br />

localization (Aki & Ferrazz<strong>in</strong>i, 2000), the <strong>envelopes</strong> provided by the scatter<strong>in</strong>g coefficients could<br />

diverge from the recorded <strong>envelopes</strong>. Two assumptions must also be made before sett<strong>in</strong>g the time<br />

soil at which diffusion is reached. The first is that, relative to scatter<strong>in</strong>g, absorption is low (Marg-<br />

er<strong>in</strong> et al., 2001). This is obviously unrealistic <strong>in</strong> a <strong>volcanic</strong> area, where <strong>in</strong>tr<strong>in</strong>sic losses due to<br />

the presence <strong>of</strong> fluids and/or melt are <strong>of</strong> importance, even if scatter<strong>in</strong>g is usually predom<strong>in</strong>ant<br />

(Del Pezzo, 2008); nevertheless, we prefer to concentrate on the loss by scatter<strong>in</strong>g, before ga<strong>in</strong><strong>in</strong>g<br />

an <strong>in</strong>sight <strong>in</strong>to <strong>in</strong>tr<strong>in</strong>sic losses. The second assumption is that strong variations <strong>in</strong> the density <strong>of</strong><br />

scatterers occur only after the time required to cross one l ∗ p,s. We will po<strong>in</strong>t out this last factor <strong>in</strong><br />

section 4.<br />

2.5 Sett<strong>in</strong>g <strong>of</strong> the s<strong>in</strong>gle scatter<strong>in</strong>g model<br />

We follow a 2D scatter<strong>in</strong>g scheme valid for both P- and S-waves, similar to the ones <strong>of</strong> Marger<strong>in</strong><br />

et al. (2000) and Przybilla et al. (2006); for P-waves, propagation is shown <strong>in</strong> Figure 1. A volcano-<br />

tectonic source emits most <strong>of</strong> its energy around 3 Hz (Chouet, 2003); hence, we consider a 3 Hz<br />

Ricker po<strong>in</strong>t source located near the center <strong>of</strong> the caldera. The impulsive source radiates totally


226<br />

227<br />

228<br />

229<br />

230<br />

231<br />

232<br />

233<br />

234<br />

235<br />

236<br />

237<br />

238<br />

239<br />

240<br />

241<br />

242<br />

243<br />

244<br />

245<br />

246<br />

12 L. De Siena E. Del Pezzo C. Thomas A. Curtis L. Marger<strong>in</strong><br />

coherent P- and S-wave energies at t = t 0 p and t = t 0 s, respectively. Wave propagation is studied<br />

us<strong>in</strong>g the scatter<strong>in</strong>g pattern def<strong>in</strong>ed <strong>in</strong> section 2.2 for different frequency bands (3 Hz and 18 Hz).<br />

The direct <strong>in</strong>tensities (I d p,s(t)) are recorded at a receiver-distance rd and at the P- and S-wave<br />

arrival times (tp,s) as:<br />

I d p,s(t) =<br />

1 rd<br />

exp(− ), (18)<br />

2πrdc0δt lp,s<br />

where δt = 0.01s is the time step <strong>of</strong> the simulation, and lp,s stands for the P- or S- mean free path.<br />

To account for the source function, we assumed a source <strong>in</strong>tensity at the source (I 0 ) sum <strong>of</strong> the P-<br />

and S-waves <strong>in</strong>tensities:<br />

I 0 (t) = 1<br />

2π δ(t − t0 p,s), (19)<br />

where t 0 p,s are the P- and S-wave enucleation times. This is just an approximation, s<strong>in</strong>ce the<br />

Green function at the receiver presents tales (THIS LAST ASSERTION MUST BE BETTER<br />

EXPLAINED IN THE NEXT DRAFT!!!!) To account for the real source shape we apply a con-<br />

volution operator (Ri(t, f)), dependent on time t and frequency f, and correspondent to a 3 Hz<br />

Ricker wavelet, to the whole envelope (Ryan, 1994):<br />

Ri(t, f) = (1 − 2πf 2 t 2 ) exp(−π 2 f 2 t 2 ), (20)<br />

with breadth (B(f), the time <strong>in</strong>terval between the center <strong>of</strong> each <strong>of</strong> its 2 characteristics side lobes):<br />

�<br />

(6)<br />

B(f) = = 0.26s (21)<br />

πf<br />

This convolution marks the source effects on the <strong>Monte</strong> <strong>Carlo</strong> solutions at every lapse time.<br />

We simulate the length <strong>of</strong> a random walk by an exponential probability low, dependent on the<br />

mean free paths (Przybilla et al., 2006). The distance between collisions (dd, Figure 1) is:<br />

dd = 1<br />

lp,s<br />

exp(− r<br />

lp,s<br />

). (22)<br />

The (eventual) change <strong>in</strong> polarization and the direction after scatter<strong>in</strong>g are triggered by the s<strong>in</strong>gle<br />

scatter<strong>in</strong>g coefficients and the relative scatter<strong>in</strong>g patterns (Figure 1, Tables 1, and Table 2). The<br />

scatter<strong>in</strong>g coefficients are a direct measure <strong>of</strong> the probability <strong>of</strong> a PP, SS, PS, or SP scatter<strong>in</strong>g<br />

(Przybilla et al., 2006). A random number def<strong>in</strong>es the polarization; a second one determ<strong>in</strong>es the<br />

direction after scatter<strong>in</strong>g (angle θ <strong>in</strong> Figure 1) us<strong>in</strong>g the lookup tables (e. g. Table 2).


247<br />

248<br />

249<br />

250<br />

251<br />

252<br />

253<br />

254<br />

255<br />

256<br />

257<br />

258<br />

259<br />

260<br />

261<br />

262<br />

263<br />

264<br />

265<br />

266<br />

267<br />

268<br />

269<br />

270<br />

271<br />

<strong>Monte</strong> <strong>Carlo</strong> <strong>seismic</strong> <strong>envelopes</strong> caldera 13<br />

After each collision, the <strong>in</strong>tensity scattered at a receiver at distance rsd (Isd) is recorded by<br />

us<strong>in</strong>g the jo<strong>in</strong>t probability <strong>of</strong> detect<strong>in</strong>g a particle <strong>of</strong> given polarization (Marger<strong>in</strong> et al., 2000):<br />

rsd exp(− lP,S<br />

Isd = P (w, a, k, θsd, κ) ∗ )<br />

. (23)<br />

The probability P is dependent on the recorded mode w (PP, SS ,PS, or SP), on θsd, the angle<br />

between the propagation direction and the scatterer- receiver direction, and on the order <strong>of</strong> the von<br />

Kármán function , κ. The dependency <strong>of</strong> P on θsd expresses the <strong>in</strong>tr<strong>in</strong>sic forward anisotropy <strong>of</strong><br />

the problem. If a particle is shot towards East at the source, the receivers <strong>in</strong> the Eastern part record<br />

<strong>in</strong>tensities much higher than the ones <strong>in</strong> the Western part, s<strong>in</strong>ce it is extremely unlikely for the<br />

particle to scatter backwards.<br />

The product <strong>of</strong> the P (or S) wave-number (kp,s = k) and the correlation length a is always larger<br />

than 1. The total number <strong>of</strong> propagated particles is N = 10 6 . The simulation can be repeated for<br />

a number <strong>of</strong> sources equivalent to the one considered <strong>in</strong> section 3. The location <strong>of</strong> the sources<br />

varies randomly <strong>in</strong> a block <strong>of</strong> 1 km side, to account for the different hypocenters. The f<strong>in</strong>al image<br />

is the average <strong>of</strong> the <strong>in</strong>tensities produced by the different sources. The code is parallelized us<strong>in</strong>g<br />

OpenMP (for the detection at different receivers) and MPI (for the simulation at different sources).<br />

This is <strong>of</strong> extreme importance at 18 Hz, due to the reduced mean free path, produc<strong>in</strong>g a large<br />

number <strong>of</strong> collisions.<br />

In Figure 3a (18 Hz) and Figure 3b (3 Hz) we show the synthetic s<strong>in</strong>gle scatter<strong>in</strong>g <strong>in</strong>tensities<br />

obta<strong>in</strong>ed for a s<strong>in</strong>gle source, κ = 0.5, and at a source-station distance<strong>of</strong> 4 km. Diffusion solutions<br />

obta<strong>in</strong>ed by us<strong>in</strong>g Equations (13), (15), and (17) for κ = 0.5 and κ = 1 (dotted l<strong>in</strong>e and plus<br />

l<strong>in</strong>e <strong>in</strong> Figure 3a,b) have the same magnitude order and follow the same exponential decay after<br />

approximately 3tS; they result <strong>in</strong>dist<strong>in</strong>guishable <strong>in</strong> logarithmic scale. We check <strong>in</strong> section 3.1 the<br />

correspondence with real data <strong>envelopes</strong> recorded at a ray-distance <strong>of</strong> 4 km.<br />

We have now a reference model to <strong>in</strong>terpret real <strong>seismic</strong> <strong>envelopes</strong> recorded <strong>in</strong> the caldera.<br />

Our aim is to understand the degree <strong>of</strong> coherency still present <strong>in</strong> early coda <strong>in</strong> different frequency<br />

bands, and to highlight the actual lapse-time at which diffusion is reached. Our assumption is<br />

rsd


272<br />

273<br />

274<br />

275<br />

276<br />

277<br />

278<br />

279<br />

280<br />

281<br />

282<br />

283<br />

284<br />

285<br />

286<br />

287<br />

288<br />

289<br />

290<br />

291<br />

292<br />

293<br />

294<br />

295<br />

296<br />

297<br />

14 L. De Siena E. Del Pezzo C. Thomas A. Curtis L. Marger<strong>in</strong><br />

that, <strong>in</strong> a <strong>volcanic</strong> environment, an unbounded model is unreliable, especially if it is exclusively<br />

triggered by a set <strong>of</strong> s<strong>in</strong>gle scatter<strong>in</strong>g l<strong>in</strong>ear coefficients.<br />

3 DATA<br />

We use the high frequency velocity record<strong>in</strong>gs <strong>of</strong> 11 low period three-component <strong>seismic</strong> stations<br />

<strong>in</strong>stalled by the University <strong>of</strong> Wiscons<strong>in</strong> <strong>in</strong>side the Campi Flegrei caldera dur<strong>in</strong>g the 1982-1984<br />

<strong>seismic</strong> crisis (e. g. Pujol & Aster (1990)). In this period, a large number <strong>of</strong> low magnitude volcano-<br />

tectonic earthquakes was located <strong>in</strong> a cube <strong>of</strong> 1 km side <strong>in</strong> the center <strong>of</strong> the caldera (G. et al., 1999).<br />

The <strong>seismic</strong>ity was relocated us<strong>in</strong>g non-l<strong>in</strong>ear algorithms, while the correspond<strong>in</strong>g waveforms<br />

were used by different authors to obta<strong>in</strong> velocity, attenuation, and scatter<strong>in</strong>g images <strong>of</strong> the caldera<br />

(De Lorenzo et al., 2001; Zollo et al., 2002; Tramelli et al., 2006; Battaglia et al., 2006; De Siena<br />

et al., 2010).<br />

We create a data set <strong>of</strong> 212 waveforms produced by 59 earthquakes located <strong>in</strong> the 1 km side<br />

cube. For each event-station pair we trace rays with the ray-bend<strong>in</strong>g approach <strong>of</strong> Block (1991) <strong>in</strong><br />

the 3D S-wave velocity structure <strong>of</strong> Battaglia et al. (2008). We then perform a 3D rotation <strong>of</strong> the<br />

<strong>seismic</strong> waveforms <strong>in</strong> the path and transverse directions, employ<strong>in</strong>g both vertical and horizontal<br />

record<strong>in</strong>gs, and compute the root-mean-square (RMS) <strong>envelopes</strong> <strong>of</strong> the seismograms <strong>in</strong> the path<br />

and transverse directions (Sato & Fehler, 1998). The transverse <strong>envelopes</strong> are the average <strong>of</strong> the<br />

two tangential components. F<strong>in</strong>ally, we shift each envelope to its S-wave arrival, and sum the traces<br />

at each station <strong>in</strong> order to form an average envelope-per-station, or array envelope.<br />

We first consider the normalized array <strong>envelopes</strong> obta<strong>in</strong>ed after filter<strong>in</strong>g <strong>in</strong> the 18 Hz and 3 Hz<br />

frequency bands, and recorded at 4 km distance (Figure 4a,b). In Figures 5-8, the correspond<strong>in</strong>g<br />

station is labeled as W02; we remark that this is not the station nearest to the epicenter area,<br />

evidenced with a X <strong>in</strong> the center <strong>of</strong> the caldera, North <strong>of</strong> Pozzuoli. In the same figures, we show<br />

the trace <strong>of</strong> the caldera rim at 2 km depth as a broad gray circumference; this signature is modeled<br />

on the velocity and scatter<strong>in</strong>g tomographic images <strong>of</strong> Battaglia et al. (2008) and Tramelli et al.<br />

(2006).


298<br />

299<br />

300<br />

301<br />

302<br />

303<br />

304<br />

305<br />

306<br />

307<br />

308<br />

309<br />

310<br />

311<br />

312<br />

313<br />

314<br />

315<br />

316<br />

317<br />

318<br />

319<br />

320<br />

321<br />

322<br />

323<br />

324<br />

325<br />

3.1 Envelopes at 4 km ray-distance<br />

<strong>Monte</strong> <strong>Carlo</strong> <strong>seismic</strong> <strong>envelopes</strong> caldera 15<br />

At 18 Hz, the synthetic coda <strong>envelopes</strong> obta<strong>in</strong>ed via the s<strong>in</strong>gle scatter<strong>in</strong>g coefficients and the array<br />

coda <strong>envelopes</strong> at station W02 are <strong>in</strong> good agreement (Figure 4a). Respect to the correspond-<br />

<strong>in</strong>g s<strong>in</strong>gle source synthetic (Figure 3a), the direct wave and the coda <strong>of</strong> the array synthetic are<br />

broadened due to the differences <strong>in</strong> source locations. The diffusion time is marked at 3tS: after<br />

this limit, late coda <strong>in</strong>tensities are <strong>of</strong> the same order <strong>of</strong> diffusion <strong>in</strong>tensities, support<strong>in</strong>g the use <strong>of</strong><br />

coda techniques. On the other hand, between 3 s and 10 s, real <strong>in</strong>tensities are twice the synthetic<br />

ones. S<strong>in</strong>gle scatter<strong>in</strong>g seems a good start<strong>in</strong>g po<strong>in</strong>t to synthesize <strong>envelopes</strong> <strong>in</strong> this frequency band.<br />

Anyway, the cause <strong>of</strong> the large <strong>in</strong>termediate-time <strong>in</strong>tensities must be <strong>in</strong>vestigated to synthesize the<br />

complete envelope.<br />

S<strong>in</strong>gle scatter<strong>in</strong>g array synthetics do not reproduce real <strong>in</strong>tensities at 3 Hz (Figure 4b). The<br />

variable source location does not model the broaden<strong>in</strong>g <strong>of</strong> coda waves nor for early nor for late<br />

coda. Data and synthetic coda present a common behavior only after 9ts. The ratio between the<br />

direct wave and the early coda <strong>in</strong>tensities is not respected as well as the phase <strong>of</strong> the conversions,<br />

absent <strong>in</strong> the synthetics. The broaden<strong>in</strong>g <strong>of</strong> the synthetic envelope is even smaller than at 18 Hz,<br />

ma<strong>in</strong>ly due to the difference <strong>in</strong> mean free paths (Table 5): a longer mean free path implies a smaller<br />

number <strong>of</strong> <strong>in</strong>teraction, and less energy arriv<strong>in</strong>g to the station.<br />

There are many possible explanations to this disagreement. First we are not consider<strong>in</strong>g a 3D<br />

model, while every tomography performed <strong>in</strong> the area evidences the 3D nature <strong>of</strong> both the attenua-<br />

tion and scatter<strong>in</strong>g wave-fields (Tramelli et al., 2006; Battaglia et al., 2008; De Siena et al., 2010).<br />

The disagreement could be due e. g. to horizontal and vertical <strong>in</strong>terfaces, like the one characteriz-<br />

<strong>in</strong>g the whole caldera at 3 km depth(e. g. Zollo et al. (2008)). We could also model the envelope<br />

us<strong>in</strong>g momentum scatter<strong>in</strong>g coefficients, as done e.g. by Yamamoto & Sato (2010). This provides<br />

a greater degree <strong>of</strong> conversion, and an <strong>in</strong>crease <strong>of</strong> backward scatter<strong>in</strong>g, therefore new phases aris-<br />

<strong>in</strong>g <strong>in</strong> the <strong>in</strong>termediate coda. A direct application <strong>of</strong> momentum scatter<strong>in</strong>g coefficient is anyway<br />

physically unfair. First, we should have more likely a <strong>boundary</strong>, caus<strong>in</strong>g the passage to a different<br />

scatter<strong>in</strong>g regime. In addition, the associated transport mean free paths (Table 5) represent more<br />

a measurement <strong>of</strong> diffusion than a characteristic scatter<strong>in</strong>g length. F<strong>in</strong>ally, momentum scatter<strong>in</strong>g


326<br />

327<br />

328<br />

329<br />

330<br />

331<br />

332<br />

333<br />

334<br />

335<br />

336<br />

337<br />

338<br />

339<br />

340<br />

341<br />

342<br />

343<br />

344<br />

345<br />

346<br />

347<br />

348<br />

349<br />

350<br />

16 L. De Siena E. Del Pezzo C. Thomas A. Curtis L. Marger<strong>in</strong><br />

coefficients completely disregard forward propagation. This is good if we model a trapp<strong>in</strong>g bound-<br />

ary; dissipation is anyway always present <strong>in</strong> real <strong>envelopes</strong> and a small amount <strong>of</strong> forward <strong>in</strong>tensity<br />

is necessary to simulate a not conservative medium.<br />

3.2 Envelopes <strong>in</strong> the caldera<br />

We will <strong>in</strong>vestigate the anomalous behavior <strong>of</strong> <strong>in</strong>termediate coda at 18 Hz, and <strong>of</strong> the whole enve-<br />

lope at 3 Hz, <strong>in</strong>clud<strong>in</strong>g the effect <strong>of</strong> determ<strong>in</strong>istic boundaries <strong>in</strong> the transport theory. Namely we<br />

assume that these effects are produced by a passage between two scatter<strong>in</strong>g regime. This change<br />

<strong>in</strong> the scatter<strong>in</strong>g texture represents anomalies <strong>of</strong> dimensions comparable with the wavelength, and<br />

enhances large angle scatter<strong>in</strong>g. Nevertheless, before deal<strong>in</strong>g with the new theory and the corre-<br />

spond<strong>in</strong>g synthetics, we will discuss the data <strong>envelopes</strong> <strong>in</strong> the whole caldera, to get a better <strong>in</strong>sight<br />

<strong>in</strong>to the scatter<strong>in</strong>g propagation at different distances.<br />

3.2.1 Envelopes at 18 Hz<br />

In Figure 5 (radial) and Figure 6 (transverse), we image the normalized <strong>envelopes</strong> recorded at<br />

each station after filter<strong>in</strong>g <strong>in</strong> the 18 Hz frequency band. Each envelope is the average <strong>of</strong> the ones<br />

produced by the 59 earthquakes <strong>in</strong> the X region. In this frequency band, we expect strong forward<br />

SS scatter<strong>in</strong>g (Table 2), produc<strong>in</strong>g a relatively high and th<strong>in</strong> peak for the S-direct arrival, with a<br />

smooth exponential decrease at larger lapse times. In terms <strong>of</strong> l<strong>in</strong>e-<strong>of</strong>-sight propagation (a topic<br />

treated <strong>in</strong> section 3.3) we expect the degree <strong>of</strong> coherence <strong>in</strong> the coda to be lower at 18 Hz than<br />

at 3 Hz; this is a consequence <strong>of</strong> the shorter mean free paths (Table 5) correspond<strong>in</strong>g to a greater<br />

number <strong>of</strong> collisions.<br />

Most <strong>of</strong> the stations <strong>in</strong> the rest <strong>of</strong> the caldera present a clear direct impulsive onset. This is<br />

followed <strong>in</strong> many cases by a second coherent phase <strong>of</strong> relatively large <strong>in</strong>tensity <strong>in</strong> the early coda<br />

(W21, W17, and, more evidently, W11 and W20). This high <strong>in</strong>tensity is particularly evident on<br />

both transverse and radial <strong>envelopes</strong> <strong>in</strong> the western and central quadrants (Figure 6, e.g. station<br />

W20). Eastern stations present <strong>in</strong>stead a larger <strong>in</strong>coherent broaden<strong>in</strong>g (e.g. compare W03 and


351<br />

352<br />

353<br />

354<br />

355<br />

356<br />

357<br />

358<br />

359<br />

360<br />

361<br />

362<br />

363<br />

364<br />

365<br />

366<br />

367<br />

368<br />

369<br />

370<br />

371<br />

372<br />

373<br />

374<br />

375<br />

<strong>Monte</strong> <strong>Carlo</strong> <strong>seismic</strong> <strong>envelopes</strong> caldera 17<br />

W04), and a s<strong>in</strong>gle coherent <strong>in</strong>tensity is hardly dist<strong>in</strong>guishable. Go<strong>in</strong>g farther from the epicenter<br />

(e.g. stations W3) coda broaden<strong>in</strong>g <strong>in</strong>creases almost everywhere.<br />

3.2.2 Envelopes at 3 Hz<br />

In Figure 7 and Figure 8, we show the normalized radial and transverse <strong>envelopes</strong> calculated <strong>in</strong><br />

the 3 Hz frequency band. We remark the differences between these <strong>envelopes</strong>, the ones calculated<br />

at 18 Hz, and the synthetic model<strong>in</strong>g (Figures 4-6). Most stations <strong>in</strong> proximity <strong>of</strong> the epicenter<br />

present a dist<strong>in</strong>ct direct peak (W17, W21, ma<strong>in</strong>ly W2), usually followed by a second peak <strong>of</strong><br />

equal or larger <strong>in</strong>tensity (e. g. station W11). The <strong>envelopes</strong> recorded North, West, and South <strong>of</strong> the<br />

epicenter (stations W4, W11, W20, and W21) present strong correlations; this common behavior<br />

proves that a similar scatter<strong>in</strong>g medium is sampled by the wave-fields. Pulse broaden<strong>in</strong>g is large<br />

everywhere, but it <strong>in</strong>creases towards East (stations W2, W3, W15), at the expenses <strong>of</strong> correlation<br />

with the other stations. The center <strong>of</strong> the caldera looks like a b<strong>in</strong>ary medium, as evidenced e.g.<br />

by scatter<strong>in</strong>g tomography (De Siena et al., 2011). As the largest ray distances, coherent phases<br />

usually disappear, and the <strong>envelopes</strong> present noise-like shape (stations W9, W10, and W14) with<br />

an exponential decrease at lapse time as large as 6 ∗ tS. The highest peak is also as late as 3 ∗ tS<br />

(e.g W5), hardly discernible from the early coda constant trend. Station W9, hav<strong>in</strong>g a high direct<br />

coherent peak at 18 Hz, is now totally noise-like, with almost no decrease with <strong>in</strong>creas<strong>in</strong>g lapse-<br />

time.<br />

3.2.3 S<strong>in</strong>gle earthquake <strong>envelopes</strong><br />

We evidence the <strong>in</strong>terference effects <strong>in</strong> coda <strong>envelopes</strong> consider<strong>in</strong>g the propagation <strong>of</strong> the wave-<br />

field produced from a s<strong>in</strong>gle earthquake, and recorded at three stations East <strong>of</strong> the epicenter (Figure<br />

9). The earthquake was recorded on the first <strong>of</strong> April 1984; both the earthquake magnitude (M = 2)<br />

and the source mechanism (shown <strong>in</strong> Figure 9) were calculated by Zollo & Bernard (1993). Hence,<br />

we remove these source effects from the unfiltered <strong>envelopes</strong> calculated at three different stations<br />

(W17, W2, and W15), disposed at <strong>in</strong>creas<strong>in</strong>g distances on a cont<strong>in</strong>uous curve. We compute the


376<br />

377<br />

378<br />

379<br />

380<br />

381<br />

382<br />

383<br />

384<br />

385<br />

386<br />

387<br />

388<br />

389<br />

390<br />

391<br />

392<br />

393<br />

394<br />

395<br />

396<br />

397<br />

398<br />

399<br />

400<br />

401<br />

402<br />

18 L. De Siena E. Del Pezzo C. Thomas A. Curtis L. Marger<strong>in</strong><br />

S-wave travel times by us<strong>in</strong>g the same ray-bend<strong>in</strong>g approach used for array analysis (Thurber &<br />

Eberhart–Phillips, 1999).<br />

With <strong>in</strong>creas<strong>in</strong>g distance, the scattered <strong>in</strong>tensity progressively <strong>in</strong>creases at the expense <strong>of</strong> the<br />

direct <strong>in</strong>tensity, and the maximum <strong>in</strong>tensity is evident at larger lapse-times. Hence, the <strong>envelopes</strong><br />

are clearly affected by broaden<strong>in</strong>g; we def<strong>in</strong>e its duration as the pulse broaden<strong>in</strong>g td, the lag time<br />

between the S wave onset and the time when the envelope decays to the half <strong>of</strong> the maximum<br />

amplitude (Saito et al., 2005). td is shown <strong>in</strong> Figure 9 as a horizontal broad gray segment on each<br />

envelope; dur<strong>in</strong>g propagation, it is always <strong>of</strong> the same order <strong>of</strong> the travel time, <strong>in</strong> contradiction with<br />

the low-angle scatter<strong>in</strong>g approximation (Gusev & Abubakirov, 1996). Filter<strong>in</strong>g at 18 Hz produces<br />

more compact <strong>envelopes</strong> (Figure 5-6), and td is usually less than the travel time; this is never true<br />

at 3 Hz.<br />

This results correlate with the difficulty found <strong>in</strong> imag<strong>in</strong>g attenuation and scatter<strong>in</strong>g at fre-<br />

quencies lower than 6 Hz (Tramelli et al., 2006; De Siena et al., 2010). The broadened <strong>envelopes</strong><br />

recorded at Campi Flegrei are an evidence <strong>of</strong> a much more complicated scatter<strong>in</strong>g pattern than<br />

the one described by a s<strong>in</strong>gle scatter<strong>in</strong>g approximation. The shape <strong>of</strong> the <strong>envelopes</strong> <strong>of</strong> Figure 9,<br />

particularly <strong>of</strong> the early coda is more similar to the one produced by a closed cavity (Derode et al.,<br />

2003), and requires at least a change <strong>in</strong> the scatter<strong>in</strong>g properties <strong>of</strong> the medium.<br />

3.3 L<strong>in</strong>e-<strong>of</strong>-sight propagation<br />

Data prove that high <strong>in</strong>tensity coherent signals trigger the <strong>envelopes</strong> at <strong>in</strong>termediate times. The un-<br />

usual broaden<strong>in</strong>g characterizes <strong>envelopes</strong> recorded <strong>in</strong> <strong>volcanic</strong> areas (Sato & Fehler, 1998; Saito<br />

et al., 2005). The high <strong>in</strong>tensity <strong>in</strong>termediate-time conversions question the opportunity <strong>of</strong> describ-<br />

<strong>in</strong>g waveform propagation by us<strong>in</strong>g the usual multiple scatter<strong>in</strong>g techniques. A description <strong>in</strong> term<br />

<strong>of</strong> l<strong>in</strong>e-<strong>of</strong>-sight propagation provides important markers to discern the latent degree <strong>of</strong> coherence<br />

<strong>in</strong> local <strong>volcanic</strong> record<strong>in</strong>gs. To f<strong>in</strong>d the cause <strong>of</strong> this unexpected coherence is the aim <strong>of</strong> this<br />

description.<br />

The total <strong>in</strong>tensity < I > we calculate as a <strong>seismic</strong> envelope represents the magnitude <strong>of</strong> the<br />

<strong>seismic</strong> wave-field, a mixture <strong>of</strong> coherent and <strong>in</strong>coherent <strong>in</strong>tensities (I c and I i , Wu (1985)). We


403<br />

404<br />

405<br />

406<br />

407<br />

408<br />

409<br />

410<br />

411<br />

412<br />

413<br />

414<br />

415<br />

416<br />

417<br />

418<br />

419<br />

420<br />

421<br />

422<br />

423<br />

424<br />

425<br />

426<br />

427<br />

<strong>Monte</strong> <strong>Carlo</strong> <strong>seismic</strong> <strong>envelopes</strong> caldera 19<br />

consider a wave <strong>in</strong>cident on the medium characterized by the ACF and PSDF described <strong>in</strong> section<br />

2.1. The field recorded at a receiver consists <strong>of</strong> the sum <strong>of</strong> the average (coherent) field (φ c 0 = 〈φ〉)<br />

and the fluctuat<strong>in</strong>g (<strong>in</strong>coherent) field (φ f<br />

0) (Ishimaru, 1997). The coherent S-wave <strong>in</strong>tensity can be<br />

expressed as:<br />

Ic = exp(−g 0 sz − bz) (24)<br />

where b is the absorption coefficient and g 0 s is given by Equation (8). Hence, the <strong>in</strong>coherent <strong>in</strong>ten-<br />

sity is:<br />

Ii = exp(−bz)[1 − exp(−g 0 sz)]. (25)<br />

We start neglect<strong>in</strong>g any <strong>in</strong>tensity contribution com<strong>in</strong>g from the region outside <strong>of</strong> the medium<br />

dimension, L (Figure 1).<br />

Apart for <strong>in</strong>tr<strong>in</strong>sic attenuation, the quantity trigger<strong>in</strong>g the ratio between coherent and <strong>in</strong>coher-<br />

ent <strong>in</strong>tensity is the optical distance ζ 0 = g 0 s ∗ z = z/ls, where ls is the mean free path. If ζ 0 1. Already at a<br />

distance <strong>of</strong> 4 km, the average ray-distance for the Campi Flegrei caldera (De Siena et al., 2010), we<br />

obta<strong>in</strong> ζ 0 > 1. The <strong>in</strong>coherent <strong>in</strong>tensity is dom<strong>in</strong>ant over the coherent one; after the direct arrival<br />

we obta<strong>in</strong> a field dom<strong>in</strong>ated by <strong>in</strong>coherent <strong>in</strong>tensities, where coherent ones are still present. In a<br />

way similar to laser propagation, 18 Hz direct waves are able to image the Earth structure crossed<br />

by the ray, s<strong>in</strong>ce only a small amount <strong>of</strong> energy is lost by backscatter<strong>in</strong>g. The ray approximation<br />

is fulfilled at first order. After the direct wave, anyway, the field is ma<strong>in</strong>ly <strong>in</strong>coherent; no coherent<br />

phases, except for the low-<strong>in</strong>tensity phases due to conversion, are present <strong>in</strong> the <strong>envelopes</strong>. The<br />

same calculation for frequency 3 Hz gives an optical distance <strong>of</strong> ζ 0 = 0.16. In this case, coherent<br />

and <strong>in</strong>coherent <strong>in</strong>tensities have comparable magnitudes (Ishimaru, 1997), and we expect coherent<br />

phases <strong>in</strong> the data <strong>envelopes</strong>.<br />

In a complicated <strong>volcanic</strong> medium like the one under study blobs <strong>of</strong> f<strong>in</strong>ite dimension diffract<br />

<strong>in</strong> a conical sector the <strong>in</strong>cident wave. At ray distance rd (e.g. Figure 1), the diffracted wave has a


428<br />

429<br />

430<br />

431<br />

432<br />

433<br />

434<br />

435<br />

436<br />

437<br />

438<br />

439<br />

440<br />

441<br />

442<br />

443<br />

444<br />

445<br />

446<br />

447<br />

448<br />

449<br />

450<br />

451<br />

452<br />

453<br />

20 L. De Siena E. Del Pezzo C. Thomas A. Curtis L. Marger<strong>in</strong><br />

f<strong>in</strong>ite spread, a measure <strong>of</strong> pulse broaden<strong>in</strong>g given by sp ∼ = 3/(g 0 s ∗ rd) (Ishimaru, 1997); as long<br />

as the spread is much less than the shadow <strong>of</strong> the blob, diffraction effects are small. Nevertheless,<br />

assum<strong>in</strong>g rd = 4 km we obta<strong>in</strong> sp ∼ = 19 at 3 Hz, and the effect <strong>of</strong> diffraction is dom<strong>in</strong>ant. At<br />

18 Hz the same calculation leads to sp ∼ = .6: as expected, diffraction effects are less important,<br />

even if still present, for high frequency waves. Incoherent <strong>in</strong>tensity will be assigned from now on<br />

to the s<strong>in</strong>gle scatter<strong>in</strong>g solutions. The coherent phases still present after the direct arrival require<br />

a change <strong>in</strong> the scatter<strong>in</strong>g texture <strong>of</strong> the medium, by the <strong>in</strong>clusion <strong>of</strong> one - or more - <strong>boundary</strong><br />

conditions <strong>in</strong> the medium.<br />

4 LARGE-ANGLE SCATTERING FROM RANDOMLY DISTRIBUTED<br />

SCATTERERS: INCLUDING DETERMINISTIC BACKSCATTERING FROM<br />

BOUNDARY<br />

Attenuation and scatter<strong>in</strong>g image Campi Flegrei caldera as a b<strong>in</strong>ary medium (e.g. Figure 5), where<br />

scatter<strong>in</strong>g <strong>in</strong>tensity at large frequencies is predom<strong>in</strong>ant <strong>in</strong> its center and borders (Tramelli et al.,<br />

2006; De Siena et al., 2011). The presence <strong>of</strong> a rim surround<strong>in</strong>g the caldera is a feasible cause for<br />

an enhancement <strong>of</strong> large angle scatter<strong>in</strong>g. This <strong>in</strong>clusion could obviously produce the determ<strong>in</strong>istic<br />

effect <strong>of</strong> a second wave-field directly produced at the <strong>boundary</strong>. In Figure X we show the synthetic<br />

<strong>seismic</strong> <strong>envelopes</strong> obta<strong>in</strong>ed add<strong>in</strong>g the reflection <strong>of</strong> each particle on the rim us<strong>in</strong>g Snell laws.<br />

It is evident that these <strong>in</strong>tensities are not relevant respect to the s<strong>in</strong>gle scatter<strong>in</strong>g solution: this<br />

is expected, s<strong>in</strong>ce each reflection happens at a different po<strong>in</strong>t <strong>of</strong> the rim, and the correspond<strong>in</strong>g<br />

<strong>in</strong>tensities are various order <strong>of</strong> magnitude less than the <strong>in</strong>coherent ones .<br />

We propose a second mechanism which could better expla<strong>in</strong> the broadened recorded en-<br />

velopes: the enhancement <strong>of</strong> the large angle scatter<strong>in</strong>g due to a drastic change <strong>in</strong> the scatter<strong>in</strong>g<br />

regime, where the caldera rim acts as a slab <strong>of</strong> randomly distributed scatterers (Ishimaru, 1997).<br />

As a first approximation, we discuss the case <strong>of</strong> normal <strong>in</strong>cidence <strong>of</strong> the particle on the rim - which<br />

is a good approximation for a source <strong>in</strong> the center, and strong forward anisotropy - and we neglect<br />

the dimension <strong>of</strong> the scatterers. We also assume that the total <strong>in</strong>tensities follow the equation <strong>of</strong>


454<br />

455<br />

456<br />

457<br />

458<br />

459<br />

460<br />

461<br />

462<br />

463<br />

464<br />

465<br />

466<br />

467<br />

468<br />

469<br />

470<br />

471<br />

472<br />

473<br />

474<br />

transfer:<br />

and cont<strong>in</strong>ue to disregard phases.<br />

<strong>Monte</strong> <strong>Carlo</strong> <strong>seismic</strong> <strong>envelopes</strong> caldera 21<br />

d<br />

ds I(r, s) = −ρσtI(r, s) + ρσt<br />

�<br />

p(s, s<br />

4π 4π<br />

′ )I(r, s ′ )dω ′ , (26)<br />

Follow<strong>in</strong>g Ishimaru (1977), we can apply the concept <strong>of</strong> zonation, where the passage from a<br />

zone <strong>of</strong> anisotropic forward scatter<strong>in</strong>g to one <strong>of</strong> large angle backward scatter<strong>in</strong>g is totally def<strong>in</strong>ed<br />

on the base <strong>of</strong> the path traveled by the particles. Namely, until each particle path is below a given<br />

soil, forward s<strong>in</strong>gle scatter<strong>in</strong>g is still dom<strong>in</strong>ant. The <strong>in</strong>clusion <strong>of</strong> a <strong>boundary</strong> allows the separation<br />

<strong>of</strong> the total <strong>in</strong>tensity <strong>in</strong> forward (I+) and backward (I−) <strong>in</strong>tensities:<br />

⎧<br />

⎪⎨ I+(r, s) s · z > 0<br />

I(r, s) =<br />

⎪⎩ I−(r, s) s · z < 0<br />

This is a major change respect to s<strong>in</strong>gle scatter<strong>in</strong>g solutions, where only conversion scatter<strong>in</strong>g<br />

could drastically change the direction from the <strong>in</strong>cident one. We start consider<strong>in</strong>g this as the unique<br />

<strong>boundary</strong> <strong>in</strong> our model, namely, an <strong>in</strong>terface separat<strong>in</strong>g the caldera - from the unbounded region<br />

outside <strong>of</strong> it (Figure Y).<br />

Equation (26) becomes a set <strong>of</strong> two equations. For the forward <strong>in</strong>tensity:<br />

d<br />

ds I+(r, s) = −ρσtI+(r, s) + ρσt<br />

�<br />

p(s, s<br />

4π<br />

′ )I+(r, s ′ )dω ′ + ρσt<br />

�<br />

4π<br />

+2π<br />

p(s, s<br />

+2π<br />

′ )I−(r, s ′ )dω ′<br />

where +2π and −2π signs stand for <strong>in</strong>tegrations for s’ <strong>in</strong> the ranges s ′ · z > 0 and s ′ · z < 0.<br />

Similarly, for s · z < 0:<br />

d<br />

ds I−(r, s) = −ρσtI−(r, s) + ρσt<br />

4π<br />

�<br />

+2π<br />

p(s, s ′ )I−(r, s ′ )dω ′ + ρσt<br />

4π<br />

�<br />

p(s, s<br />

+2π<br />

′ )I+(r, s ′ )dω ′<br />

Forward and backward <strong>in</strong>tensities are connected; <strong>in</strong> the high frequency case, multiple scatter<strong>in</strong>g<br />

forward <strong>in</strong>tensity should be dom<strong>in</strong>at<strong>in</strong>g, with a small contribution com<strong>in</strong>g from backward. Instead,<br />

<strong>in</strong> presence <strong>of</strong> a slab separat<strong>in</strong>g two media hav<strong>in</strong>g different scatter<strong>in</strong>g features - determ<strong>in</strong>istic <strong>in</strong><br />

nature - equation <strong>of</strong> transport - which is <strong>in</strong>stead statistical - can be transformed.<br />

We highlighted the small differences caused by a determ<strong>in</strong>istic reflection on the <strong>boundary</strong> <strong>in</strong><br />

Figure X. With the new approach, each time the propagation distance <strong>of</strong> a particle overcomes the<br />

source-station <strong>boundary</strong>, we use the transport equation for I− given by (29). The forward <strong>in</strong>tensity<br />

(27)<br />

(28)<br />

(29)


475<br />

476<br />

477<br />

478<br />

479<br />

480<br />

481<br />

482<br />

483<br />

484<br />

485<br />

486<br />

487<br />

488<br />

489<br />

490<br />

491<br />

492<br />

493<br />

494<br />

495<br />

496<br />

497<br />

498<br />

499<br />

500<br />

22 L. De Siena E. Del Pezzo C. Thomas A. Curtis L. Marger<strong>in</strong><br />

I+ <strong>in</strong> last factor is given by the equation (28) computed at the previous scatter<strong>in</strong>g event, without<br />

the contribution <strong>of</strong> backward <strong>in</strong>tensity. Afterwards, the propagation will still be modeled by us<strong>in</strong>g<br />

the transport equation <strong>in</strong> the form <strong>of</strong> Equations (27), (28), and (29), with the forward <strong>in</strong>tensity.<br />

4.1 The complete large-angle bounded <strong>Monte</strong> <strong>Carlo</strong> simulation<br />

The problem <strong>of</strong> model<strong>in</strong>g <strong>in</strong>termediate coda <strong>in</strong>tensities for a caldera is approached constra<strong>in</strong><strong>in</strong>g<br />

part <strong>of</strong> the medium with a circular <strong>boundary</strong> condition: this is similar to <strong>in</strong>clud<strong>in</strong>g a circular rim <strong>of</strong><br />

ray 5 km, and center at the epicenter <strong>in</strong> the s<strong>in</strong>gle scatter<strong>in</strong>g simulation. The stations can be <strong>in</strong>side<br />

the rim-<strong>boundary</strong>, <strong>in</strong> the proximity (on) the rim-<strong>boundary</strong>, or far-outside <strong>of</strong> it.<br />

We can both:<br />

• model a totally-dissipat<strong>in</strong>g <strong>boundary</strong>, where I− is not created after cross<strong>in</strong>g the <strong>boundary</strong>,<br />

• model a mixed reflective-dissipat<strong>in</strong>g <strong>boundary</strong>, and consider the I− contribution after cross<strong>in</strong>g<br />

the <strong>boundary</strong>.<br />

The presence <strong>of</strong> a totally-dissipative <strong>boundary</strong> does not produce a backward <strong>in</strong>tensity: its effect<br />

is to dissipate the forward <strong>in</strong>tensity only, and the equation <strong>of</strong> transport must not be updated. We<br />

model dissipation by us<strong>in</strong>g momentum scatter<strong>in</strong>g coefficients and transport mean free paths, <strong>in</strong>-<br />

stead <strong>of</strong> completely disregard<strong>in</strong>g forward <strong>in</strong>tensity after the <strong>boundary</strong>: this will have consequences<br />

on the reach <strong>of</strong> diffusion and on the correlation <strong>of</strong> late coda synthetics with data.<br />

In the case <strong>of</strong> a mixed reflective and dissipat<strong>in</strong>g <strong>boundary</strong>, the <strong>in</strong>terface does not produce a<br />

determ<strong>in</strong>istic wave-field - an effective reflection phase at each scatter<strong>in</strong>g triggered by Snell’s laws.<br />

Instead, each time the particle-source distance becomes larger than the source-<strong>boundary</strong> distance,<br />

a source produc<strong>in</strong>g energy <strong>in</strong> a random backward direction on the <strong>boundary</strong> itself is added to the<br />

<strong>in</strong>tensities recorded at the different stations. The backward angle is triggered by the momentum<br />

scatter<strong>in</strong>g coefficients - s<strong>in</strong>ce they exclude the forward direction. Afterwards, Equations (27) - (29)<br />

trigger propagation.<br />

We can start consider<strong>in</strong>g the iteration solution n - eventually, the solution at scatter<strong>in</strong>g event n<br />

- after cross<strong>in</strong>g the <strong>boundary</strong>. From this po<strong>in</strong>t, Equation (29) rules backward <strong>in</strong>tensity, and its first


501<br />

502<br />

503<br />

504<br />

505<br />

506<br />

507<br />

508<br />

509<br />

510<br />

511<br />

512<br />

513<br />

514<br />

515<br />

516<br />

517<br />

518<br />

519<br />

<strong>Monte</strong> <strong>Carlo</strong> <strong>seismic</strong> <strong>envelopes</strong> caldera 23<br />

contribution I− is dependent on the (n − 1)th iteration solution <strong>of</strong> I+ - I+(n−1):<br />

d<br />

ds I−(r, s) = −ρσtI−(r, s) + ρσt<br />

�<br />

p(s, s<br />

4π<br />

′ )I−(r, s ′ )dω ′ + ρσt<br />

�<br />

4π<br />

+2π<br />

p(s, s<br />

+2π<br />

′ )I+(n−1)(r, s ′ )dω ′ .<br />

For the forward first iteration after the <strong>boundary</strong>, I+(n−1), we can assume I−n−1 = I−0 = 0 <strong>in</strong><br />

Equation (28) - s<strong>in</strong>ce backward propagation starts only after cross<strong>in</strong>g the slab - and we cont<strong>in</strong>ue<br />

to consider the usual transport equation:<br />

d<br />

ds I+n(r, s) = −ρσtI+n(r, s) + ρσt<br />

4π<br />

�<br />

(30)<br />

p(s, s<br />

+2π<br />

′ )I+n(r, s ′ )dω ′ + 0. (31)<br />

Nevertheless, we model dissipation us<strong>in</strong>g the transport mean free path, and the contribution <strong>of</strong> I+<br />

after the <strong>boundary</strong> ma<strong>in</strong>ly affects late coda.<br />

The first backward <strong>in</strong>tensity recorded results dependent on forward <strong>in</strong>tensity: this last can be<br />

considered as the source <strong>of</strong> backward <strong>in</strong>tensity rewrit<strong>in</strong>g Equation (30) as:<br />

d<br />

ds I−(r, s) + ρσtI−(r, s) − ρσt<br />

�<br />

p(s, s<br />

4π<br />

′ )I−(r, s ′ )dω ′ = ρσt<br />

�<br />

4π<br />

+2π<br />

p(s, s<br />

+2π<br />

′ )I+(n−1)(r, s ′ )dω ′ .<br />

This corresponds to the s<strong>in</strong>gle scatter<strong>in</strong>g solution for a statistical source created by the <strong>boundary</strong>.<br />

The contribution <strong>of</strong> the simple <strong>boundary</strong> can be calculated as the probability <strong>of</strong> record<strong>in</strong>g the<br />

backscattered <strong>in</strong>tensity at the n forward-<strong>in</strong>tensity scatter<strong>in</strong>g. We rewrite the detected <strong>in</strong>tensity <strong>of</strong><br />

equation (23) as forward <strong>in</strong>tensity at the n scatter<strong>in</strong>g (I n+<br />

sd ):<br />

I n+<br />

sd = KP (w, a, k, θsd,<br />

rsd exp(−<br />

κ) ∗<br />

rsd<br />

(32)<br />

lP,S )<br />

, (33)<br />

where the different factors are still dependent on the s<strong>in</strong>gle scatter<strong>in</strong>g coefficients. After cross<strong>in</strong>g<br />

the <strong>boundary</strong>, we add a second <strong>in</strong>tensity (I (n−1) −sd):<br />

I −<br />

sd = KP ∗ (w, a, k, θsd,<br />

rsd exp(−<br />

κ) ∗<br />

rsd<br />

lP,S )<br />

, (34)<br />

Our first hypothesis is that P* is still dependent on s<strong>in</strong>gle scatter<strong>in</strong>g coefficients after the <strong>in</strong>ter-<br />

action. This corresponds to a s<strong>in</strong>gle scatter<strong>in</strong>g solution - an <strong>in</strong>tensity that propagates <strong>in</strong> the same<br />

medium, but with a large angle deviation respect to I+ (Figure X). Nevertheless the particles -<br />

which have a certa<strong>in</strong> size distribution and some absorption - produce a multiple scatter<strong>in</strong>g - not a<br />

s<strong>in</strong>gle scatter<strong>in</strong>g - <strong>in</strong>tensity. Physically multiple scatter<strong>in</strong>g backscatter<strong>in</strong>g <strong>in</strong>tensity is not depen-


520<br />

521<br />

522<br />

523<br />

524<br />

525<br />

526<br />

527<br />

528<br />

529<br />

530<br />

531<br />

532<br />

533<br />

534<br />

535<br />

536<br />

24 L. De Siena E. Del Pezzo C. Thomas A. Curtis L. Marger<strong>in</strong><br />

dent anymore on the scatter<strong>in</strong>g cross-section - therefore on the s<strong>in</strong>gle scatter<strong>in</strong>g coefficients - but<br />

on the power absorbed by the particle, the absorption cross-section σa (Ishimaru, 1997). The dif-<br />

ference can be understood consider<strong>in</strong>g a slab <strong>of</strong> particles hav<strong>in</strong>g a different scatter<strong>in</strong>g regime - not<br />

a simple <strong>boundary</strong> separat<strong>in</strong>g the caldera from an unbounded region.<br />

4.2 Slab<br />

Ishimaru (1997) formulates and gives analytic solutions to the previous <strong>in</strong>tensity equations for a<br />

wave-field perpendicularly <strong>in</strong>cident on a slab - therefore on a region hav<strong>in</strong>g a f<strong>in</strong>ite dimension<br />

d - <strong>in</strong> the small angle approximation (Figure Z). Introduc<strong>in</strong>g longitud<strong>in</strong>al (z) and transverse (ρ)<br />

coord<strong>in</strong>ates, and rewrit<strong>in</strong>g numerical density as ρn:<br />

∂<br />

∂z I−(z, ρ, s) + s · ∇tI−(z, ρ, s) = −ρnσtI−(z, ρ, s) + ρnσt<br />

4π<br />

where<br />

Q = ρnσt<br />

4π<br />

�<br />

� � +∞<br />

−∞<br />

p(s − s ′ )I−(z, ρ, s ′ )dω ′ + Q.<br />

(35)<br />

p(s − s<br />

+2π<br />

′ )I+(z, ρ, s ′ )dω ′ . (36)<br />

Q is the source term for the backward simulation - like the one due to a simple <strong>boundary</strong> - but now<br />

a second <strong>in</strong>tensity dependent on d must be added at the receiver.<br />

The analytic solution can be obta<strong>in</strong>ed if we consider that:<br />

p(s − s ′ ) = σb<br />

σt<br />

, (37)<br />

where σb is the backscatter<strong>in</strong>g cross section. We can also write I+(z) as the attenuated <strong>in</strong>tensity at<br />

the <strong>boundary</strong> for a plane wave hav<strong>in</strong>g <strong>in</strong>tensity I0 at the source:<br />

I+(z) = I0 exp(−ρnσaB), (38)<br />

where B is the source-<strong>boundary</strong> distance. Substitut<strong>in</strong>g Equations (37) and (38) <strong>in</strong> Equation (36)<br />

we obta<strong>in</strong> the new source term:<br />

Q = ρnσb<br />

4π I0 exp(−ρnσaB) exp(−ρnσaz). (39)


537<br />

538<br />

539<br />

540<br />

541<br />

542<br />

543<br />

544<br />

545<br />

546<br />

547<br />

548<br />

549<br />

550<br />

551<br />

552<br />

553<br />

554<br />

555<br />

556<br />

557<br />

558<br />

559<br />

<strong>Monte</strong> <strong>Carlo</strong> <strong>seismic</strong> <strong>envelopes</strong> caldera 25<br />

The analytical solution <strong>of</strong> Equation (35) with the new source term is given by (Ishimaru, 1997):<br />

I−(z, s) = ρnσbI0d [1 − exp(−2ρnσad)]<br />

exp(−ρnσaB). (40)<br />

4π 2ρnσad<br />

The s<strong>in</strong>gle scatter<strong>in</strong>g backward solution is <strong>in</strong>stead:<br />

I f<br />

−(z, s) = ρnσbI0d<br />

4π<br />

[1 − exp(−2ρnσtd)]<br />

2ρnσtd<br />

exp(−ρnσaB). (41)<br />

This relations are applicable <strong>in</strong> the range θ > (kp,sl T P,S )− 1: for S-waves this corresponds to θ >??<br />

or phenomena like backscatter<strong>in</strong>g enhancement should be modeled (Marger<strong>in</strong> & van Tiggelen,<br />

2001; van Tiggelen et al., 2001). Nevertheless, these phenomena are excluded from our tractation,<br />

s<strong>in</strong>ce our source-detection distance is always larger than one elastic wavelength at both frequen-<br />

cies.<br />

It is evident that, <strong>in</strong> presence <strong>of</strong> a multiple scatter<strong>in</strong>g regime, backward scatter<strong>in</strong>g from s<strong>in</strong>gle<br />

scatter<strong>in</strong>g solution underestimates real <strong>in</strong>tensity, and that absorption must be added to get the<br />

correct ones.<br />

5 RESULTS<br />

5.1 Results at 18 Hz<br />

The number <strong>of</strong> <strong>in</strong>teraction before the <strong>boundary</strong> is large <strong>in</strong> this case, and the forward <strong>in</strong>tensity<br />

lost after cross<strong>in</strong>g it is low. Dissipation after the <strong>boundary</strong> is modeled by us<strong>in</strong>g the momentum<br />

scatter<strong>in</strong>g coefficients and transport mean free path - which are <strong>of</strong> the order <strong>of</strong> more than 100<br />

km (Table 4). At 18 Hz (Figure 10a) this is not equivalent to completely disregard<strong>in</strong>g scatter<strong>in</strong>g<br />

propagation after the <strong>boundary</strong>. Most <strong>of</strong> the scattered energy is produced before: nevertheless, late<br />

coda <strong>in</strong>tensities <strong>in</strong>crease due to the contribution <strong>of</strong> the momentum scatter<strong>in</strong>g coefficients (gray<br />

dotted l<strong>in</strong>e). The ratio between direct and coda <strong>in</strong> real data can only be reached summ<strong>in</strong>g these to<br />

the s<strong>in</strong>gle scatter<strong>in</strong>g <strong>in</strong>tensities (black dotted l<strong>in</strong>e).<br />

The presence <strong>of</strong> a dissipative <strong>boundary</strong> caus<strong>in</strong>g large angle scatter<strong>in</strong>g does not produce the<br />

arrival <strong>of</strong> a second <strong>in</strong>tensity - or, better, a general broaden<strong>in</strong>g <strong>of</strong> the envelope <strong>in</strong> the <strong>in</strong>termediate<br />

coda (Figure 10a). This happens consider<strong>in</strong>g a reflect<strong>in</strong>g <strong>boundary</strong> - Figure 10b.


560<br />

561<br />

562<br />

563<br />

564<br />

565<br />

566<br />

567<br />

568<br />

569<br />

570<br />

571<br />

572<br />

573<br />

574<br />

575<br />

576<br />

577<br />

578<br />

579<br />

580<br />

581<br />

582<br />

583<br />

584<br />

585<br />

26 L. De Siena E. Del Pezzo C. Thomas A. Curtis L. Marger<strong>in</strong><br />

NOT PRESENT IN FIGURE 12!!!<br />

5.2 Results at 3 Hz<br />

The result <strong>of</strong> the model<strong>in</strong>g with a dispersive caldera at 3 Hz is shown <strong>in</strong> Figure 11a, for the station<br />

at 4 km <strong>in</strong>side the rim. The <strong>in</strong>clusion <strong>of</strong> momentum scatter<strong>in</strong>g dissipation allows fora relevant<br />

<strong>in</strong>crease <strong>of</strong> the late coda: this confirms their importance for model<strong>in</strong>g low-frequency coda even for<br />

volcano-tectonic earthquakes (Sato et al., 2004).<br />

THIS SIMULATION MUST BE UPDATED!!!<br />

The peak near the S-direct arrival is due to the contribution <strong>of</strong> I n+ <strong>in</strong> equation (33). The<br />

envelope modeled without consider<strong>in</strong>g I n+ after the cross<strong>in</strong>g the <strong>boundary</strong> is shown <strong>in</strong> Figure Yb.<br />

STILL TO BE DONE!!!!<br />

A second wave-field at time almost equal to twice the travel time <strong>of</strong> the direct wave is created<br />

by the <strong>in</strong>teraction with the <strong>boundary</strong>.<br />

6 DISCUSSION<br />

Attenuation tomography has been performed with the ray method, and suggests the reliability <strong>of</strong><br />

a ray approximation at these frequencies (De Siena et al., 2010). At 3 Hz the average wavelength<br />

is <strong>of</strong> the order <strong>of</strong> the correlation length <strong>of</strong> the area, as estimated by De Siena et al. (2011). This<br />

creates large <strong>in</strong>stabilities <strong>in</strong> the <strong>seismic</strong> coda, even at large lapse-time, which do not allow reliable<br />

attenuation and scatter<strong>in</strong>g images (Tramelli et al., 2006; De Siena et al., 2010). These anomalies<br />

are evidently produced by the <strong>in</strong>teraction <strong>of</strong> the wave-field with the medium <strong>in</strong> the Mie scatter<strong>in</strong>g<br />

regime: they can therefore provide new h<strong>in</strong>ts on the structure <strong>of</strong> the caldera.<br />

The forward <strong>in</strong>tensity after cross<strong>in</strong>g the <strong>boundary</strong> is ruled by momentum scatter<strong>in</strong>g coeffi-<br />

cients, <strong>in</strong> the case <strong>of</strong> dissipat<strong>in</strong>g caldera. It is therefore quickly lost due to the dimension <strong>of</strong> the<br />

transport mean free path (Tables 4). We do not change the propagation reference system to pre-<br />

serve total <strong>in</strong>tensity. Backward <strong>in</strong>tensity (I − ) is still ruled by the mean free path: what makes the<br />

difference between the two frequency ranges is the ratio between this last quantity (around .16 km<br />

for 18 Hz and 5 km for 3Hz) and the source-<strong>boundary</strong> distance (5 km). At 18 Hz the <strong>in</strong>teraction


586<br />

587<br />

588<br />

589<br />

590<br />

591<br />

592<br />

593<br />

594<br />

595<br />

596<br />

597<br />

598<br />

599<br />

600<br />

601<br />

602<br />

603<br />

604<br />

605<br />

606<br />

607<br />

608<br />

609<br />

610<br />

611<br />

612<br />

613<br />

<strong>Monte</strong> <strong>Carlo</strong> <strong>seismic</strong> <strong>envelopes</strong> caldera 27<br />

with the <strong>boundary</strong> happens after several mean free paths, and the wave-field is triggered by the<br />

scatter<strong>in</strong>g regime <strong>in</strong>side the caldera. At 3 Hz the <strong>in</strong>teraction with the <strong>boundary</strong> is cont<strong>in</strong>uous, and<br />

the wave-field suffers too many <strong>in</strong>teractions to become diffusive at the expected time-soil. In both<br />

cases the effect <strong>of</strong> a <strong>boundary</strong> is seen clearly at <strong>in</strong>termediate times: at 3 Hz anyway if affects also<br />

the late coda, forbidd<strong>in</strong>g the diffusion reach.<br />

Synthetic and data examples show that, if a determ<strong>in</strong>istic approach provides unrealistic mod-<br />

els <strong>of</strong> waveform <strong>envelopes</strong>, a totally stochastic one is also unreliable, if <strong>boundary</strong> conditions are<br />

not taken <strong>in</strong>to account. As the frequency <strong>in</strong>creases, momentum scatter<strong>in</strong>g coefficients are able to<br />

model late coda <strong>in</strong>tensities. Diffusion solutions prove the efficiency <strong>of</strong> diffusion-derived methods<br />

- as coda-normalization - for times greater than 3tS <strong>in</strong> this frequency range. At <strong>in</strong>termediate times<br />

anyway, the effect <strong>of</strong> the <strong>boundary</strong> is also <strong>of</strong> importance <strong>in</strong> this frequency range, while a determ<strong>in</strong>-<br />

istic location <strong>of</strong> larger scale structure by a direct application <strong>of</strong> Snell laws results unfeasible.<br />

At 3 Hz the passage to a different scatter<strong>in</strong>g regime and <strong>of</strong> the <strong>boundary</strong> is highlighted <strong>in</strong> the<br />

data by the recursive coherence at almost constant <strong>in</strong> the <strong>in</strong>termediate and late coda. We model<br />

these evidences <strong>in</strong>troduc<strong>in</strong>g three spatial quantities: mean and transport mean free path, and dis-<br />

tance between source and <strong>boundary</strong>. The correlation between synthetic and data <strong>envelopes</strong> is strik-<br />

<strong>in</strong>g, even <strong>in</strong> the 2D simulation. This is actually not surpris<strong>in</strong>g: the rim considered <strong>in</strong> the simulation<br />

can be easily transformed <strong>in</strong> an horizontal <strong>in</strong>terface - as the one generally present under Campi<br />

Flegrei (Zollo et al., 2008; Battaglia et al., 2008; De Siena et al., 2010). The only parameter to<br />

provide results the source-<strong>boundary</strong> distance, both <strong>in</strong> the vertical and horizontal directions.<br />

De Siena et al. (2010) and De Siena et al. (2011) recently highlighted the differences <strong>in</strong> at-<br />

tenuation and scatter<strong>in</strong>g characteriz<strong>in</strong>g the eastern and western part <strong>of</strong> the Campi Flegrei caldera.<br />

The eastern part is dom<strong>in</strong>ated by <strong>in</strong>tr<strong>in</strong>sic attenuation, with its center approximately located <strong>in</strong><br />

the zone <strong>of</strong> Solfatara. A <strong>seismic</strong> <strong>in</strong>terface is located below Campi Flegrei at around 3 km depth,<br />

as evidenced by reflection seismology (Zollo et al., 2008); attenuation tomography reveals that<br />

this <strong>in</strong>terface is broken by an high attenuation anomaly <strong>in</strong> its center (black region <strong>in</strong> Figures 5-8).<br />

In the central and western parts, scatter<strong>in</strong>g attenuation is much stronger than <strong>in</strong>tr<strong>in</strong>sic (De Siena<br />

et al., 2011). We expect that the differences between real and synthetic <strong>envelopes</strong> could be reduced


614<br />

615<br />

616<br />

617<br />

618<br />

619<br />

620<br />

621<br />

622<br />

623<br />

624<br />

625<br />

626<br />

627<br />

628<br />

629<br />

630<br />

631<br />

632<br />

633<br />

634<br />

635<br />

636<br />

637<br />

638<br />

639<br />

640<br />

28 L. De Siena E. Del Pezzo C. Thomas A. Curtis L. Marger<strong>in</strong><br />

model<strong>in</strong>g the determ<strong>in</strong>istic effect <strong>of</strong> these scatter<strong>in</strong>g anomalies, and that, contrary to what happens<br />

for high frequency late coda, first-order coda decrease might be heavily <strong>in</strong>fluenced by large-angle<br />

scatter<strong>in</strong>g at lower frequencies<br />

7 CONCLUSIONS<br />

We demonstrate by means <strong>of</strong> <strong>Monte</strong> <strong>Carlo</strong> simulation and array process<strong>in</strong>g <strong>of</strong> real <strong>envelopes</strong> that<br />

a statistical approach cannot deal with the description <strong>of</strong> whole <strong>seismic</strong> <strong>envelopes</strong> <strong>in</strong> an active<br />

caldera, unless determ<strong>in</strong>istic <strong>in</strong>terfaces are <strong>in</strong>cluded <strong>in</strong>to the scatter<strong>in</strong>g medium. We sketched this<br />

<strong>in</strong>terfaces start<strong>in</strong>g from a simplistic model - ma<strong>in</strong>ly <strong>in</strong>troduc<strong>in</strong>g a circular <strong>boundary</strong>, as well as<br />

a l<strong>in</strong>ear drastic change <strong>in</strong>to the scatter<strong>in</strong>g properties <strong>of</strong> the caldera. The effect is anyway statis-<br />

tical and can be dealt with transport theory - namely with the <strong>in</strong>clusion <strong>of</strong> large-angle scatter<strong>in</strong>g<br />

enhanced by <strong>in</strong>terfaces, and careful use <strong>of</strong> dissipation. We implemented <strong>in</strong> a s<strong>in</strong>gle scatter<strong>in</strong>g sim-<br />

ulation both effects through the direct application <strong>of</strong> backward transport theory (Ishimaru, 1997).<br />

We def<strong>in</strong>e the scattered <strong>in</strong>tensities through the def<strong>in</strong>ition <strong>of</strong> first-order and momentum scatter<strong>in</strong>g<br />

coefficients - this last are used to account for the dissipation outside the caldera rim.<br />

This model is still applicable when the fractional fluctuation are small, and successful for<br />

higher-frequency bands because <strong>of</strong> self-similarity <strong>of</strong> the media spectra; however, as noted, the<br />

characteristic time <strong>of</strong> the Markov approximation <strong>in</strong>creases to the order <strong>of</strong> the travel time even at 18<br />

Hz, testify<strong>in</strong>g not only the <strong>in</strong>adequacy <strong>of</strong> s<strong>in</strong>gle scatter<strong>in</strong>g solutions, but also <strong>of</strong> the diffusion model<br />

at lower frequencies. The <strong>in</strong>termediate <strong>envelopes</strong> <strong>in</strong> the frequency ranges <strong>in</strong>vestigated cannot be<br />

modeled without the addition <strong>of</strong> determ<strong>in</strong>istic effects, caus<strong>in</strong>g enhanced large angle <strong>in</strong>tensities.<br />

The passage to a complete multiple scatter<strong>in</strong>g synthetic model <strong>in</strong> this <strong>in</strong>terval results necessary.<br />

The Markov approximation has been so far the ma<strong>in</strong> road to add diffraction effects to the<br />

s<strong>in</strong>gle scattered wave-field <strong>in</strong> order to describe envelope broaden<strong>in</strong>g and maximum amplitude<br />

decay. We approach the problem <strong>in</strong>clud<strong>in</strong>g coherency <strong>in</strong>formation directly <strong>in</strong>to the <strong>Monte</strong> <strong>Carlo</strong><br />

simulation, without the requirement for additional propagators. The envelope demonstrate the <strong>need</strong><br />

for a momentum scatter<strong>in</strong>g coefficient to model late coda <strong>envelopes</strong> recorded <strong>in</strong> an active caldera,<br />

and the persistence <strong>of</strong> the coherency for the first seconds after the P- and S-wave onsets. Diffusion


641<br />

642<br />

643<br />

644<br />

645<br />

646<br />

647<br />

648<br />

649<br />

650<br />

651<br />

652<br />

653<br />

654<br />

655<br />

656<br />

657<br />

658<br />

659<br />

660<br />

661<br />

662<br />

663<br />

664<br />

665<br />

<strong>Monte</strong> <strong>Carlo</strong> <strong>seismic</strong> <strong>envelopes</strong> caldera 29<br />

is usually reached after a lapse time <strong>of</strong> 10 s (3 times the S-wave travel time) at 4 km for 18 Hz, as<br />

demonstrated by the comparison <strong>of</strong> synthetics, real data, and diffusion solutions.<br />

Model<strong>in</strong>g a volcano-tectonic earthquake envelope - namely, the energy-with-time produced<br />

by an impulsive source - even at small source receiver distance, requires a great computational<br />

effort. It can lead to direct measures <strong>of</strong> the attenuation properties <strong>of</strong> the medium, and, at a second<br />

step, to their imag<strong>in</strong>g through tomography (Tramelli et al., 2006; De Siena et al., 2010). These are<br />

important issues for the assessment <strong>of</strong> <strong>volcanic</strong> risk and <strong>volcanic</strong> eruption monitor<strong>in</strong>g; they give<br />

the possibility to locate melt and feed<strong>in</strong>g systems trough their effect on <strong>seismic</strong> waveforms. In the<br />

future, the effect <strong>of</strong> boundaries and slab will have to be added to usual tomography techniques, to<br />

get a major <strong>in</strong>sight <strong>in</strong>to the effect <strong>of</strong> <strong>in</strong>terference on attenuation and scatter<strong>in</strong>g images.<br />

The <strong>in</strong>terference effect <strong>of</strong> a second wave-field can be used as marker to understand complex<br />

processes which could have consequences on the application <strong>of</strong> coda wave <strong>in</strong>terferometry (Snieder<br />

et al., 2002). Anyway, a complete Markov FD simulation will be necessary, not to rely on char-<br />

acteristic time to account for diffraction effects. In addition, <strong>Monte</strong> <strong>Carlo</strong> model<strong>in</strong>g with complex<br />

determ<strong>in</strong>istic scatter<strong>in</strong>g pattern - e. g. circular and elliptic r<strong>in</strong>g - could give new h<strong>in</strong>t on the gener-<br />

ation <strong>of</strong> this <strong>in</strong>terference effects for real calderas.<br />

F<strong>in</strong>ally, the observation <strong>of</strong> a dissipative range at long ray distances, and the strong <strong>in</strong>crease <strong>of</strong><br />

coherent <strong>in</strong>tensity caused by larger scale anomalies suggest the application <strong>of</strong> turbulence theory<br />

to the model<strong>in</strong>g <strong>of</strong> coda <strong>envelopes</strong>. The equations model<strong>in</strong>g the wave <strong>in</strong>tensity for a medium<br />

characterized by 2 different scales <strong>of</strong> heterogeneities - namely eddys and blobs - could provide an<br />

unprecedented view on the topic <strong>of</strong> <strong>volcanic</strong> coda wave model<strong>in</strong>g.<br />

ACKNOWLEDGMENTS<br />

The HPC-EUROPA project provided the necessary resources for the development <strong>of</strong> the program:<br />

We thank the whole staff at EPCC (Ed<strong>in</strong>burgh Parallel Comput<strong>in</strong>g Center) <strong>in</strong> Ed<strong>in</strong>burgh, and<br />

particularly Dr. Adam Carter, who were fundamental for the parallelization <strong>of</strong> the code.


666<br />

667<br />

668<br />

669<br />

670<br />

671<br />

672<br />

673<br />

674<br />

675<br />

676<br />

677<br />

678<br />

679<br />

680<br />

681<br />

682<br />

683<br />

684<br />

685<br />

686<br />

687<br />

688<br />

689<br />

690<br />

691<br />

692<br />

693<br />

30 L. De Siena E. Del Pezzo C. Thomas A. Curtis L. Marger<strong>in</strong><br />

References<br />

Aki, K. & Ferrazz<strong>in</strong>i, V., 2000. Seismic monitor<strong>in</strong>g and model<strong>in</strong>g <strong>of</strong> an active volcano for pre-<br />

diction, Journal <strong>of</strong> Geophysical Research, 105, 16617–16640.<br />

Battaglia, J., Troise, C., Obrizzo, F., P<strong>in</strong>gue, F., & Natale, G. D., 2006. Evidence <strong>of</strong> fluid migra-<br />

tion as the source <strong>of</strong> deformation at campi flegrei caldera (italy), Geophysical Research Letters.,<br />

33(L01307), doi:10.1029/2005GL024904.<br />

Battaglia, J., Zollo, A., Virieux, J., & Dello Iacono, D., 2008. Merg<strong>in</strong>g active and passive data sets<br />

<strong>in</strong> traveltime tomography: the case study <strong>of</strong> Campi Flegrei caldera (Southern Italy), Geophysical<br />

Prospect<strong>in</strong>g, 56, 555–573.<br />

Block, L. V., 1991. Jo<strong>in</strong>t hypocenter–velocity <strong>in</strong>version <strong>of</strong> local earthquake arrival time data <strong>in</strong><br />

two geothermal regions, Ph.d. dissertation, Massachusetts Institute <strong>of</strong> Technology, Cambridge.<br />

Brenguier, F., Shapiro, N. M., Campillo, M., Ferrazz<strong>in</strong>i, V., Duputel, Z., Coutant, O., & Nerces-<br />

sian, A., 2008. Towards forecast<strong>in</strong>g <strong>volcanic</strong> eruptions us<strong>in</strong>g <strong>seismic</strong> noise, Nature Geoscience,<br />

January, doi :10.1038/ngeo104.<br />

Chouet, B., 2003. Volcano seismology, PAGEOPH, 160, 739–788.<br />

De Lorenzo, S., Zollo, A., & Mongelli, F., 2001. Source parameters and three-dimensional at-<br />

tenuation structure from the <strong>in</strong>version <strong>of</strong> microearthquake pulse width data: Qp imag<strong>in</strong>g and<br />

<strong>in</strong>ferences on the thermal state <strong>of</strong> the Campi Flegrei caldera (southern Italy), Journal <strong>of</strong> Geo-<br />

physical Research, 106, 16265–16286.<br />

De Siena, L., Del Pezzo, E., & Bianco, F., 2010. Campi Flegrei <strong>seismic</strong> attenuation image:<br />

evidences <strong>of</strong> gas resevoirs, hydrotermal bas<strong>in</strong>s and feed<strong>in</strong>g systems, Journal <strong>of</strong> Geophysical<br />

Research, 115(B0), 9312–9329.<br />

De Siena, L., Del Pezzo, E., & Bianco, F., 2011. A scatter<strong>in</strong>g image <strong>of</strong> Campi Flegrei from the<br />

autocorrelation functions <strong>of</strong> velocity tomograms, Geophysical Journal International, 184(3),<br />

1304–1310.<br />

Del Pezzo, E., 2008. Seismic wave scatter<strong>in</strong>g <strong>in</strong> volcanoes, <strong>in</strong> Earth Heterogeneity and Scatter<strong>in</strong>g<br />

Effects <strong>of</strong> Seismic Waves, vol. 50 <strong>of</strong> Advances <strong>in</strong> Geophysics, chap. 13, pp. 353–369, Elsevier,<br />

Amsterdam.


694<br />

695<br />

696<br />

697<br />

698<br />

699<br />

700<br />

701<br />

702<br />

703<br />

704<br />

705<br />

706<br />

707<br />

708<br />

709<br />

710<br />

711<br />

712<br />

713<br />

714<br />

715<br />

716<br />

717<br />

718<br />

719<br />

720<br />

721<br />

<strong>Monte</strong> <strong>Carlo</strong> <strong>seismic</strong> <strong>envelopes</strong> caldera 31<br />

Derode, A., Larose, E., Tanter, M., de Rosny, J., Tour<strong>in</strong>, A., Campillo, M., & F<strong>in</strong>k, M., 2003.<br />

Recover<strong>in</strong>g the green’s function from field-field correlations <strong>in</strong> an open scatter<strong>in</strong>g medium (l),<br />

Journal <strong>of</strong> the Acoustic Society <strong>of</strong> America, 113(6), 2973–2976.<br />

Fehler, M. C., Sato, H., & Huang, L. J., 2000. Envelope broaden<strong>in</strong>g <strong>of</strong> outgo<strong>in</strong>g waves <strong>in</strong> 2d<br />

random media: a comparison between the markov approximation and numerical simulations,<br />

Bullet<strong>in</strong> <strong>of</strong> the Seismological Society <strong>of</strong> America, 90, 914–928.<br />

Fukushima, Y., Nishizawa, O., Sato, H., & Ohtake, M., 2003. Laboratory study on scatter<strong>in</strong>g<br />

characteristics <strong>of</strong> shear waves <strong>in</strong> rock samples, Bullet<strong>in</strong> <strong>of</strong> the Seismological Society <strong>of</strong> America,<br />

93(1), 253–263.<br />

G., O., Civetta, L., Gaudio, C. D., de Vita, S., Vito, M. A. D., Isaia, R., Petrazzuoli, S. M., Ric-<br />

ciardi, G., & Ricco, C., 1999. Short-term ground deformations and <strong>seismic</strong>ity <strong>in</strong> the resurgent<br />

Campi Flegrei caldera (Italy): an example <strong>of</strong> active block-resurgence <strong>in</strong> a densely populated<br />

area, <strong>in</strong> Volcanism <strong>in</strong> the Campi Flegrei, vol. 91, pp. 415 – 451, eds G., O., Civetta, L., &<br />

Valent<strong>in</strong>e, G. A., Elsevier.<br />

Gusev, A. A. & Abubakirov, I. R., 1987. <strong>Monte</strong>-carlo simulation <strong>of</strong> record envelope <strong>of</strong> a near<br />

earthquake, Physics <strong>of</strong> the Earth and Planetary Interiors, 49, 30–36.<br />

Gusev, A. A. & Abubakirov, I. R., 1996. Simulated <strong>envelopes</strong> <strong>of</strong> non-isotropically scattered body<br />

waves as compared to observed ones: Another manifestation <strong>of</strong> fractal heterogeneity, Geophys-<br />

ical Journal International, 127, 49–60.<br />

Henn<strong>in</strong>o, R., Tregueres, N., Shapiro, N. M., Marger<strong>in</strong>, L., van Tiggelen, B. A., & Weaver, R. L.,<br />

2001. Observation <strong>of</strong> Equipartition <strong>of</strong> Seismic Waves, Physical Review Letters, 86(15), 3447–<br />

3450.<br />

Hoshiba, M., 1991. Simulation <strong>of</strong> multiple–scattered coda wave excitation based on the energy<br />

conservation law, Physics <strong>of</strong> the Earth and Planetary Interiors, 67, 123–136.<br />

Ishimaru, A., 1977. Theory and Application <strong>of</strong> Wave Propagation and Scatter<strong>in</strong>g <strong>in</strong> Random<br />

Media, Proceed<strong>in</strong>gs <strong>of</strong> the IEEE, 65(7), 1030 – 1061.<br />

Ishimaru, A., 1997. Wave Propagation and Scatter<strong>in</strong>g <strong>in</strong> Random Media, The IEEE/OUP Series<br />

on Electromagnetic Wave Theory, Oxford University Press and IEEE Press, Oxford, 2nd edn.


722<br />

723<br />

724<br />

725<br />

726<br />

727<br />

728<br />

729<br />

730<br />

731<br />

732<br />

733<br />

734<br />

735<br />

736<br />

737<br />

738<br />

739<br />

740<br />

741<br />

742<br />

743<br />

744<br />

745<br />

746<br />

747<br />

748<br />

749<br />

32 L. De Siena E. Del Pezzo C. Thomas A. Curtis L. Marger<strong>in</strong><br />

K., H. & Levander, A. R., 1992. A stochastic view <strong>of</strong> lower crustal fabric based on evidence from<br />

the Ivrea zone, Geophysical Research Letters, 19, 1153 – 1156.<br />

Marger<strong>in</strong>, L., 2006. Attenuation, transport and diffusion <strong>of</strong> scalar waves <strong>in</strong> textured media,<br />

Tectonophysics, 416, 229–244.<br />

Marger<strong>in</strong>, L. & Nolet, G., 2003. Multiple scatter<strong>in</strong>g <strong>of</strong> high-frequency <strong>seismic</strong> waves <strong>in</strong> the deep<br />

Earth: Model<strong>in</strong>g and numerical examples, Journal <strong>of</strong> Geophysical Research, 108(B5), 2234–<br />

2249.<br />

Marger<strong>in</strong>, L. & van Tiggelen, M. C. B. A., 2001. Coherent backscatter<strong>in</strong>g <strong>of</strong> acoustic waves <strong>in</strong><br />

the near field, Geophysical Journal International, 145, 593–603.<br />

Marger<strong>in</strong>, L., Campillo, M., & Tiggelen, B. V., 1998. Radiative transfer and diffusion <strong>of</strong> waves <strong>in</strong><br />

a layered medium: New <strong>in</strong>sight <strong>in</strong>to coda Q, Geophysical Journal International, 134, 596–612.<br />

Marger<strong>in</strong>, L., Campillo, M., & van Tiggelen, B. A., 2000. <strong>Monte</strong> <strong>Carlo</strong> simulation <strong>of</strong> multiple<br />

scatter<strong>in</strong>g <strong>of</strong> elastic waves, Journal <strong>of</strong> Geophysical Research, 105(B4), 7873–7892.<br />

Marger<strong>in</strong>, L., van Tiggelen, B. A., & Campillo, M., 2001. Effect <strong>of</strong> absorption on energy parti-<br />

tion<strong>in</strong>g <strong>of</strong> elastic waves <strong>in</strong> the <strong>seismic</strong> coda, Bullett<strong>in</strong> <strong>of</strong> the Seismological Society <strong>of</strong> America,<br />

91, 624 – 627.<br />

Przybilla, J. & Korn, M., 2008. <strong>Monte</strong> carlo simulation <strong>of</strong> radiative energy transfer <strong>in</strong> cont<strong>in</strong>u-<br />

ous elastic random media - three-component <strong>envelopes</strong> and numerical validation, Geophysical<br />

Journal International, 173, 566–576.<br />

Przybilla, J., Korn, M., & Wegler, U., 2006. Radiative transfer <strong>of</strong> elastic waves versus f<strong>in</strong>ite<br />

difference simulations <strong>in</strong> two-dimensional random media, Journal <strong>of</strong> Geophysical Research,<br />

111(B0), 4305–4317.<br />

Przybilla, J., Wegler, U., & Korn, M., 2009. Estimation <strong>of</strong> crustal scatter<strong>in</strong>g parameters with<br />

elastic radiative transfer theory, Journal <strong>of</strong> Geophysical Research, 111(B0), 4305–4317.<br />

Pujol, W. S. & Aster, R., 1990. Jo<strong>in</strong>t hypocentral determ<strong>in</strong>ation and the detection <strong>of</strong> low-velocity<br />

anomalies. an example from the phlegrean fields earthquakes, Bullet<strong>in</strong> <strong>of</strong> the Seismological<br />

Society <strong>of</strong> America, 80(1), 129–139.<br />

Ryan, H., 1994. Ricker, Ormsby, Klauder, Butterworth - A Choice <strong>of</strong> Wavelets, Canadian Society


750<br />

751<br />

752<br />

753<br />

754<br />

755<br />

756<br />

757<br />

758<br />

759<br />

760<br />

761<br />

762<br />

763<br />

764<br />

765<br />

766<br />

767<br />

768<br />

769<br />

770<br />

771<br />

772<br />

773<br />

774<br />

775<br />

776<br />

777<br />

<strong>of</strong> Exploration Geophysicist Recorder, 19(7), 8 – 9.<br />

<strong>Monte</strong> <strong>Carlo</strong> <strong>seismic</strong> <strong>envelopes</strong> caldera 33<br />

Rytov, S. M., Kravtsov, Y. A., & Tatarskii, V. I., 1987. Pr<strong>in</strong>ciples <strong>of</strong> Statistical Radiophysics, vol.<br />

4, Wave Propagation Through Random Media, Spr<strong>in</strong>ger-Verlag, New York.<br />

Ryzhik, L., Papanicolaou, G., & Keller, J. B., 1996. Transport equations for elastic and other<br />

waves <strong>in</strong> random media, Wave motion, 24, 327 – 370.<br />

Saito, T., Sato, H., & Ohtake, M., 2002. Envelope broaden<strong>in</strong>g <strong>of</strong> spherically outgo<strong>in</strong>g waves <strong>in</strong><br />

three-dimensional random media hav<strong>in</strong>g power law spectra, Journal <strong>of</strong> Geophysical Research,<br />

107(B5), 2089–2103.<br />

Saito, T., Sato, H., Ohtake, M., & Obara, K., 2005. Unified explanation <strong>of</strong> envelope broaden<strong>in</strong>g<br />

and maximumamplitude decay <strong>of</strong> high–frequency seismograms based on the envelope simula-<br />

tion us<strong>in</strong>g the markov approximation: Forearc side <strong>of</strong> the <strong>volcanic</strong> front <strong>in</strong> northeastern honshu,<br />

japan, Journal <strong>of</strong> Geophysical Research, 110(B0), 1304.<br />

Sanchez-Sesma, F., 1993. Topographic effects for <strong>in</strong>cident P, SV and Raileigh waves, Tectono-<br />

physics, 218(1–3), 113–125.<br />

Sato, H., 1989. Broaden<strong>in</strong>g <strong>of</strong> seismogram <strong>envelopes</strong> <strong>in</strong> the randomly <strong>in</strong>homogeneous litho-<br />

sphere based on the parabolic approximation: southeastern honshu, japan, Journal <strong>of</strong> Geophys-<br />

ical Research, 94, 17735–17747.<br />

Sato, H. & Fehler, M. C., 1998. Seismic Wave Propagation and Scatter<strong>in</strong>g <strong>in</strong> the heterogeneous<br />

Earth, Spr<strong>in</strong>ger and Verlag, New York, USA.<br />

Sato, H., Fehler, M., & Saito, T., 2004. Hybrid synthesis <strong>of</strong> scalar wave <strong>envelopes</strong> <strong>in</strong> two-<br />

dimensional random media hav<strong>in</strong>g rich short-wavelength spectra, Journal <strong>of</strong> Geophysical Re-<br />

search, 109(B0), 6303–6314.<br />

Shapiro, S. A. & Hubral, P., 1999. Elastic Waves <strong>in</strong> Random Media, Fundamentals <strong>of</strong> Seismic<br />

Stratigraphic Filter<strong>in</strong>g, Spr<strong>in</strong>ger, Berl<strong>in</strong>, Germany.<br />

Shiomi, K., Sato, H., & Ohtake, M., 1997. Broad–band power–law spectra <strong>of</strong> well–log data <strong>in</strong><br />

japan, Geophysical Journal International, 130, 57–64.<br />

Snieder, R., Grêt, A., , Doumba, H., & Scales, J., 2002. Coda wave <strong>in</strong>terferometry for estimat<strong>in</strong>g<br />

nonl<strong>in</strong>ear behavior <strong>in</strong> <strong>seismic</strong> velocity, Science, 295, 455–473.


778<br />

779<br />

780<br />

781<br />

782<br />

783<br />

784<br />

785<br />

786<br />

787<br />

788<br />

789<br />

790<br />

791<br />

792<br />

793<br />

794<br />

795<br />

796<br />

797<br />

798<br />

799<br />

800<br />

801<br />

802<br />

803<br />

804<br />

805<br />

34 L. De Siena E. Del Pezzo C. Thomas A. Curtis L. Marger<strong>in</strong><br />

Thurber, C. H. & Eberhart–Phillips, D., 1999. Local earthquake tomography with flexible grid-<br />

d<strong>in</strong>g, Computers and Geosciences, 25, 809–818.<br />

Tramelli, A., Pezzo, E. D., Bianco, F., & Boschi, E., 2006. 3d scatter<strong>in</strong>g image <strong>of</strong> the campi<br />

flegrei caldera (southern italy). new h<strong>in</strong>ts on the position <strong>of</strong> the old caldera rim, Physics <strong>of</strong> the<br />

Earth and Planetary Interiors, 155, 269–280.<br />

Turner, J. A., 1998. Scatter<strong>in</strong>g and Diffusion <strong>of</strong> Seismic Waves, Bullett<strong>in</strong> <strong>of</strong> the Seismological<br />

Society, 88(1), 276–283.<br />

van Tiggelen, B. A., Marger<strong>in</strong>, L., & Campillo, M., 2001. Coherent backscatter<strong>in</strong>g <strong>of</strong> elastic<br />

waves: Specific role <strong>of</strong> source, polarization, and nearfield, Journal <strong>of</strong> the Acoustic Society <strong>of</strong><br />

America, 110(3), 1291–1298.<br />

Weaver, R. L., 1990. Diffusivity <strong>of</strong> ultrasound <strong>in</strong> polycrystals, Journal <strong>of</strong> the Mechanics and<br />

Physics <strong>of</strong> Solids, 1(38), 55–86.<br />

Wegler, U., 2003. Analysis <strong>of</strong> Multiple Scatter<strong>in</strong>g at Vesuvius Volcano, Italy, us<strong>in</strong>g Data <strong>of</strong> the<br />

TomoVes active <strong>seismic</strong> experiment, Journal <strong>of</strong> Volcanology and Geothermal Research, 128,<br />

45–63.<br />

Wegler, U. & Lühr, B. G., 2001. Scatter<strong>in</strong>g behaviour at merapi volcano (java) revealed from an<br />

active <strong>seismic</strong> experiment, Geophysical Journal International, 145, 579–592.<br />

Wu, R. S., 1985. Multiple scatter<strong>in</strong>g and energy transfer <strong>of</strong> <strong>seismic</strong> waves - separation <strong>of</strong> sact-<br />

ter<strong>in</strong>g effect for <strong>in</strong>tr<strong>in</strong>sic attenuation, i, Geophysical Journal <strong>of</strong> the Royal Astronomical Society,<br />

82, 57–80.<br />

Yamamoto, M. & Sato, H., 2010. Multiple scatter<strong>in</strong>g and mode conversion revealed by an active<br />

<strong>seismic</strong> experiment at asama volcano, japan, Journal <strong>of</strong> Geophysical Research, 115(B0), 7304–<br />

7317.<br />

Yoshimoto, K., 2000. <strong>Monte</strong> carlo simulation <strong>of</strong> seismogram <strong>envelopes</strong> <strong>in</strong> scatter<strong>in</strong>g media,<br />

Journal <strong>of</strong> Geophysical Research, B3(105), 6153–6162.<br />

Zeng, Y., 1991. Compact solution for multiple scattered wave energy <strong>in</strong> time doma<strong>in</strong>, Bullet<strong>in</strong> <strong>of</strong><br />

the Seismological Society <strong>of</strong> America, 81, 1022–1029.<br />

Zeng, Y., 1993. Theory <strong>of</strong> scattered p– and s–wave energy <strong>in</strong> a random isotropic scatter<strong>in</strong>g


806<br />

807<br />

808<br />

809<br />

810<br />

811<br />

812<br />

813<br />

814<br />

815<br />

816<br />

817<br />

818<br />

819<br />

820<br />

821<br />

822<br />

823<br />

824<br />

825<br />

826<br />

827<br />

828<br />

829<br />

830<br />

831<br />

<strong>Monte</strong> <strong>Carlo</strong> <strong>seismic</strong> <strong>envelopes</strong> caldera 35<br />

medium, Bullet<strong>in</strong> <strong>of</strong> the Seismological Society <strong>of</strong> America, 83, 1264–1276.<br />

Zeng, Y., Su, F., & Aki, K., 1991. Scatter<strong>in</strong>g wave energy propagation <strong>in</strong> a random isotropic<br />

scatter<strong>in</strong>g medium i. theory, Journal <strong>of</strong> Geophysical Research, 96, 607–619.<br />

Zollo, A. & Bernard, P., 1993. Fault mechanisms from near-source data: jo<strong>in</strong>t <strong>in</strong>version <strong>of</strong> S<br />

polarizations and P polarities Seismic Evidence for a Low–Velocity Zone <strong>in</strong> the Upper Crust<br />

Beneath Mount Vesuvius, Geophysical Journal International, 104, 441–451.<br />

Zollo, A., D’Auria, L., Matteis, R. D., andJ. Virieux, A. H., & Gaspar<strong>in</strong>i, P., 2002. Bayesian<br />

estimation <strong>of</strong> 2-d p-velocity models from active <strong>seismic</strong> arrival time data: imag<strong>in</strong>g <strong>of</strong> the shallow<br />

structure <strong>of</strong> mt vesuvius (southern italy), Geophysical Journal International, 151, 566–582.<br />

Zollo, A., Maerckl<strong>in</strong>, N., Vassallo, M., Dello Iacono, D., Virieux, J., & Gaspar<strong>in</strong>i, P., 2008. Seis-<br />

mic reflections reveal a massive melt layer feed<strong>in</strong>g Campi Flegrei caldera, Geophysical Re-<br />

search Letters, 35(L112306), doi:10.1029/2008GL034242.<br />

8 FIGURE LEGENDS<br />

Figure 1: Sketch <strong>of</strong> the s<strong>in</strong>gle scatter<strong>in</strong>g simulation for a P-wave source (big black square) produc-<br />

<strong>in</strong>g direct <strong>in</strong>tensity (Id) and scattered <strong>in</strong>tensities (Isd) <strong>in</strong> a medium <strong>of</strong> average velocity V0. At each<br />

collision on scatterers (small black squares), the scattered <strong>in</strong>tensity is recorded at each station (tri-<br />

angles). This <strong>in</strong>tensity is dependent both on the scatter<strong>in</strong>g angle, θsd, and on the scatterer-receiver<br />

distance, rsd. rd is the source-station distance, while dd is the distance between scatterers. A fi-<br />

nite probability <strong>of</strong> PS (and SP) conversion exists; <strong>in</strong> this case, a PS conversion deviates strongly<br />

deviates the particle from its <strong>in</strong>cidence direction.<br />

Figure 2: Power spectral density functions (PSDFs) <strong>of</strong> two-dimensional von Kármántype ran-<br />

dom media used <strong>in</strong> this study.<br />

Figure 3: Time dependence <strong>of</strong> the diffusion equations for κ = 0.5 (dotted l<strong>in</strong>e) κ = 1 (plus<br />

l<strong>in</strong>e) at 18 Hz (a) and 3 Hz (b) over-imposed on the result <strong>of</strong> the first-order <strong>Monte</strong>-<strong>Carlo</strong> simulation<br />

for a s<strong>in</strong>gle source (black cont<strong>in</strong>uous l<strong>in</strong>e).<br />

Figure 4: Transverse <strong>envelopes</strong> (gray l<strong>in</strong>e) recorded at station W02 <strong>in</strong> the 18 Hz (a) and 3Hz


832<br />

833<br />

834<br />

835<br />

836<br />

837<br />

838<br />

839<br />

840<br />

841<br />

842<br />

843<br />

844<br />

845<br />

846<br />

847<br />

848<br />

849<br />

850<br />

851<br />

36 L. De Siena E. Del Pezzo C. Thomas A. Curtis L. Marger<strong>in</strong><br />

(b) frequency bands are compared with the correspond<strong>in</strong>g s<strong>in</strong>gle scatter<strong>in</strong>g synthetics (black gray<br />

l<strong>in</strong>e).<br />

Figure 5: Normalized radial <strong>envelopes</strong> obta<strong>in</strong>ed at 18 Hz <strong>in</strong> the Campi Flegrei caldera. The<br />

X represents the area <strong>of</strong> the epicenters (a block <strong>of</strong> 1 km side). The results <strong>of</strong> the separation <strong>of</strong><br />

<strong>in</strong>tr<strong>in</strong>sic and scatter<strong>in</strong>g attenuation performed by De Siena et al. (2011) is over-imposed <strong>in</strong> the<br />

center. Light gray and black represent high scatter<strong>in</strong>g attenuation and high <strong>in</strong>tr<strong>in</strong>sic attenuation,<br />

respectively. The dotted broad circumference is a simplified model <strong>of</strong> the caldera rim at 1.5 km<br />

deduced by scatter<strong>in</strong>g and velocity images (Tramelli et al., 2006; Battaglia et al., 2008).<br />

Figure 6: Same as Figure 5 for the transverse component.<br />

Figure 7: Same as Figure 5 for the 3 Hz frequency band.<br />

Figure 8: Same as Figure 5 for the transverse component and the 3 Hz frequency band.<br />

Figure 9: Map <strong>of</strong> the Campi Flegrei caldera with three unfiltered <strong>envelopes</strong> recorded at stations<br />

W17, W02, and W15. The epicenter is <strong>in</strong> the 1 km block <strong>of</strong> Figures 5-8; its focal mechanism was<br />

used to correct the envelope <strong>in</strong>tensities, us<strong>in</strong>g the results <strong>of</strong> Zollo & Bernard (1993). The gray<br />

circle (caldera rim) is also shown.<br />

Figure 10: Sketch <strong>of</strong> the simulation with caldera rim as <strong>boundary</strong>.<br />

Figure 11: Envelope recorded at station W02 with the result <strong>of</strong> the simulation at 18 Hz with a<br />

dissipative caldera (cont<strong>in</strong>uous black l<strong>in</strong>e). Intensities produced by propagation <strong>in</strong>side the caldera<br />

rim are shown with a black dotted l<strong>in</strong>e. After the reach<strong>in</strong>g the <strong>boundary</strong>, the total <strong>in</strong>tensity is<br />

triggered by the transport mean free path (gray dotted l<strong>in</strong>e).


Figure 1.<br />

<strong>Monte</strong> <strong>Carlo</strong> <strong>seismic</strong> <strong>envelopes</strong> caldera 37


38 L. De Siena E. Del Pezzo C. Thomas A. Curtis L. Marger<strong>in</strong><br />

Figure 2.<br />

Table 1. S<strong>in</strong>gle scatter<strong>in</strong>g coefficients at different frequencies and for different von Kármán orders<br />

κ = 0.5<br />

κ = 1<br />

3 Hz 18 Hz<br />

gss 0.03819 1.61805<br />

gpp 0.00776 0.31248<br />

gps 0.00205 0.00233<br />

gsp 0.00090 0.00102<br />

gss 0.06334 2.56472<br />

gpp 0.01260 0.49416<br />

gps 0.00109 0.00022<br />

gsp 0.00048 0.00010


Intensity<br />

Intensity<br />

x 10 -4<br />

3<br />

2<br />

1<br />

0<br />

x 10 -5<br />

4<br />

3<br />

2<br />

1<br />

0<br />

5<br />

5<br />

10 15<br />

Time (s)<br />

10 15<br />

Time (s)<br />

Figure 3.<br />

<strong>Monte</strong> <strong>Carlo</strong> <strong>seismic</strong> <strong>envelopes</strong> caldera 39<br />

20<br />

20<br />

25<br />

25<br />

a)<br />

30<br />

b)<br />

30


40 L. De Siena E. Del Pezzo C. Thomas A. Curtis L. Marger<strong>in</strong><br />

Figure 4.


Figure 5.<br />

<strong>Monte</strong> <strong>Carlo</strong> <strong>seismic</strong> <strong>envelopes</strong> caldera 41


42 L. De Siena E. Del Pezzo C. Thomas A. Curtis L. Marger<strong>in</strong><br />

Figure 6.


Figure 7.<br />

<strong>Monte</strong> <strong>Carlo</strong> <strong>seismic</strong> <strong>envelopes</strong> caldera 43


44 L. De Siena E. Del Pezzo C. Thomas A. Curtis L. Marger<strong>in</strong><br />

Figure 8.<br />

Figure 9.


Figure 10.<br />

Figure 11.<br />

<strong>Monte</strong> <strong>Carlo</strong> <strong>seismic</strong> <strong>envelopes</strong> caldera 45


46 L. De Siena E. Del Pezzo C. Thomas A. Curtis L. Marger<strong>in</strong><br />

Table 2. Angular distribution <strong>of</strong> SS (first column) and PS (second) scatter<strong>in</strong>g at 3 Hz. The angle is randomly<br />

selected, and def<strong>in</strong>es the deviation from the <strong>in</strong>cident direction after scatter<strong>in</strong>g.<br />

Random number SS angle (rad) PS angle (rad)<br />

1 0.0010 0.1274<br />

2 0.0020 0.1626<br />

3 0.0030 0.1880<br />

4 0.0040 0.2088<br />

5 0.0050 0.2227<br />

6 0.0060 0.2432<br />

7 0.0072 0.2582<br />

... ... ...<br />

97 0.5998 2.6570<br />

98 0.7798 2.7222<br />

99 1.2875 2.8124<br />

100 2.2496 3.1378<br />

101 3.1411 3.1415<br />

102 4.0336 3.1454<br />

103 4.9957 3.4708<br />

104 5.5034 3.5610<br />

105 5.6802 3.6262<br />

... ... ...<br />

106 6.0260 6.0250<br />

195 6.2772 6.0400<br />

196 6.2782 6.0605<br />

197 6.2792 6.0744<br />

198 6.2802 6.0952<br />

199 6.2812 6.1206<br />

200 6.2822 6.1558<br />

201 6.2832 6.2832


<strong>Monte</strong> <strong>Carlo</strong> <strong>seismic</strong> <strong>envelopes</strong> caldera 47<br />

Table 3. Momentum scatter<strong>in</strong>g coefficients at different frequencies and for different von Kármán orders<br />

κ = 0.5<br />

κ = 1<br />

3 Hz 18 Hz<br />

g m ss 0.00222 0.00370<br />

g m pp 0.00180 0.00335<br />

g m ps 0.00136 0.00143<br />

g m sp 0.00060 0.00063<br />

g m ss 0.00103 0.00136<br />

g m pp 0.00111 0.00139<br />

g m ps 0.00046 0.00008<br />

g m sp 0.00020 0.00004<br />

Table 4. Cos<strong>in</strong>e weighted scatter<strong>in</strong>g coefficients at different frequencies and for different von Kármán orders<br />

κ = 0.5<br />

κ = 1<br />

3 Hz 18 Hz<br />

gss 0.03610 1.61305<br />

gpp 0.00704 0.31007<br />

gps 0.00119 0.00140<br />

gsp 0.00052 0.00061<br />

gss 0.06186 2.56211<br />

gpp 0.01185 0.49266<br />

gps 0.00072 0.00015<br />

gsp 0.00031 0.00007<br />

Table 5. Mean and transport mean free paths, and diffusion constants obta<strong>in</strong>ed at 3 Hz and 18 Hz for κ = .5.<br />

(km) 3 Hz 18 Hz<br />

lp 101.94 3.18<br />

ls 25.58 0.62<br />

l m p 316.02 208.98<br />

l m s 354.99 231.12<br />

l ∗ p 545.17 268.02<br />

l ∗ s 427.98 193.62<br />

D m 404.50 262.77<br />

D 559.50 262.47

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!