A Guide to the Seismic

Wavefield as seen by DAS

BRIAN L.N. KENNETT

Research School of Earth Sciences

Australian National University
A Guide to the Seismic
Wavefield as seen by DAS

BRIAN L.N. KENNETT

Research School of Earth Sciences

The Australian National University
Brian L.N. Kennett is Emeritus Professor of Seismology at the Australian
National University. His research interests are directed towards understanding
the structure of the Earth from seismological observations. He is the recipient
of numerous awards and medals for his work and is a Fellow of the Australian
Academy of Sciences and the Royal Society (London). He is the author of more
than 340 research papers and ten other books.

©2024 Brian Kennett

This work is licensed under CC BY 4.0

Contents

Preface page v
1 Introduction 1
1.1 DAS applications 1
1.2 Structure of the work 3
2 Optical fibre sensing 5
2.1 Interaction of laser pulses with fibre 5
2.2 DAS exploiting Rayleigh scattering 7
2.2.1 Relation of DAS signal to strain rate 8
2.2.2 Processing issues 10
2.2.3 Cable coupling 11
2.2.4 Channel calibration 11
3 Representing the seismic wavefield for DAS 12
3.1 Strain rate and ground velocity 12
3.2 Building the wavefield 14
3.2.1 Structural response 14
3.2.2 Constructing synthetic records 16
3.3 DAS - strain rate and orientation 18
3.3.1 Local field 18
3.3.2 Distant field 18
3.3.3 Orientation factors 21
3.4 Gauge length issues 23
3.5 Gauge length effects for a bent cable 27
4 Modal fields 31
4.1 Surface waves 31
4.2 The nature of the modal field 33
4.3 Interacting waveguides 36
4.4 Multi-mode surface waves and leaking modes 37
4.5 Modal spectra from DAS recording 40
5 Near field 43
5.1 Urban noise environment 44
5.2 Taptest effects 46

iii
iv Contents

5.3 Local wavefields 50

5.4 Comparison of in-line and offset sources 52
5.4.1 In-line propagation 55
5.4.2 Offset shots 56
5.4.3 Source effects 58
6 Regional wavefield 61
6.1 Near-regional regime 61
6.1.1 DAS observations 63
6.1.2 Simulating the regional DAS field 66
6.1.3 DAS studies of regional seismic sources 70
6.2 Far-regional to near-teleseismic regime 70
7 Global wavefield 73
7.1 The influence of DAS recording 73
7.2 Teleseismic observations 75
7.3 Simulating the teleseismic DAS field 79
7.3.1 Body waves 79
7.3.2 Surface waves 83
8 Ambient noise correlation 84
8.1 Strain correlation 85
8.2 Visualising cross-correlation effects 87
8.3 DAS orientation effects 89
8.4 Rayleigh and Love waves from DAS correlation 92
8.5 Mapping local heterogeneity 94
9 DAS array response 97
9.1 DAS array stacks 97
9.2 DAS arrays for correlation analysis 105
10 Exploiting the DAS wavefield directly 110
Bibliography 112
Preface

The short work is designed to provide an introduction to the character of the

seismic wavefield as seen with DAS recording on surface cables. It often makes
use of representations of the wavefield in the frequency – slowness domain with
allowance for the orientation of the cable with respect to the passing seismic
waves. Here I build on results from my earlier books The Seismic Wavefield;
Volumes I & II (Kennett, 2001, 2002).
For DAS records an important role is played by the finite gauge length
employed to extract the signals relating to change in optical path length. This
means that results at a designated channel are not point measurements but
rather averaged versions of strain-rate. For local events the dominant part of
the strain rate can be extracted from the difference of ground velocity resolved
along the fibre at the ends of the gauge interval. For more distant events, the
response at seismic frequencies can be represented as the acceleration along
the fibre modulated by the wave slowness resolved in the same direction, which
means there is a strong dependence on cable orientation. In the examples of
different aspects of the seismic wavefield as recorded by DAS I have drawn on
my long experience in the analysis of seismograms to try to point out the ways
in which uni-directional strain(rate) modifies the relative importance of seismic
phases.
Slowness-domain representation of the DAS signal also allows analysis of the
array response of cable configurations, indicating a bias due to the slowness
weighting associated with the effect of gauge length. Unlike seismometer arrays
the response is not described by a single generic stacking function.
When I first became interested in the potential of Distributed Acoustic
Sensing (DAS) in early 2020 I found it difficult to get a clear overview of the
principles and so started to develop my own tools to understand the way
in which the dense but uni-directional sampling of DAS records renders the
seismic wavefield in near-surface recordings. This led to the treatment in
Kennett (2022) which is augmented here. Recent publications have helped
to provide an improved background (e.g., Lindsey & Martin, 2021). A good
summary of the concepts behind the DAS approach is provided in Part I of

v
vi Preface

the monograph edited by Li, Karrenbach & Ajo-Franklin (2022). There has
been a dramatic growth in studies exploiting DAS and a extensive survey of the
literature from 2012 – 2023 is provided by F. Cheng (2024).
Much of the code used in this work is available at the AuPass repository
Research page: [Link] Fortran
codes for surface wave dispersion frequency-slowness (phase speed) synthesis
and display are made available at Zenodo (Kennett, 2023b, doi: 10.5281/zen-
odo.7384357).

Acknowledgements
I was introduced to DAS issues by Meghan Miller and Andreas Fichtner and I
am grateful to them for continuing interactions. In particular, the challenges
provided by DAS recording on the Greenland ice-cap have encouraged careful
thinking about the DAS specific issues.
Several groups have provided material for the range of observational
examples: Andreas Fichtner, Daniel Bowden and Krystyna Smolinski (ETH
Zürich); Voon Hui Lai, Meghan Miller and Herb McQueen (RSES, ANU); Erdinc
Saygin and Ludovic Ricard (CSIRO, Perth).
The AusPass repository is enabled by AuScope ([Link]) via the
National Collaborative Research Infrastructure Strategy (NCRIS) of the
Australian Government.
1
Introduction

Distributed Acoustic Sensing is one of a variety of techniques that have evolved

to exploit the interaction of laser light with optical fibres. In the presence of
a dynamic disturbance the properties of the fibre are subtly modified and this
affects the way that light propagates along the fibre. Variations in the properties
of backscattered light from minor imperfections in the fibre can be used to map
out the passage of waves past the fibre.
With repetitive sampling by laser pulses, interferometric sensing can be used
to measure the phase of the backscattered light as a function of time from the
emission of the pulse. The position along the fibre can then be found from the
two-way transit time. The relative phase of the scattered energy along the fibre
as a function of time provides a measure of local strain or strain rate, and so a
single fibre can become a large suite of effective sensors (e.g., Parker et al., 2014;
Zhan 2020). The use of such multiple measurements of strain or strain rate
extracted from Rayleigh backscattering has been termed distributed acoustic
sensing (DAS). Although DAS results are presented as time series at specific
points, the nature of the sampling process means that there is averaging over
a ‘gauge length’ around the nominal point. The opportunities provided by the
dense sampling of the wavefield by DAS have seen a wide range of geophysical
applications and a rapidly growing community of users.

1.1 DAS applications

The initial developments for DAS came in exploration geophysics with borehole
installations (e.g., Mestayer et al., 2011; Daley et al., 2016; Karrenbach et al.,
2018). Such dense sampling also offers considerable potential for surface
reflection profiling (e.g., Bakulin et al., 2020), but the lack of broadside
sensitivity limits the exploitation of steeply travelling reflected P waves unless
special cable configurations are used (e.g., Kuvshinov, 2016; Wuestefeld & Wicks,
2019).
In recent years there have been a wide range of DAS applications in surface
seismology (see, e.g., Lindsey & Martin, 2021), with a broadening range of
studies as more groups move into the field. Event monitoring has been carried

1
2 Introduction

out in a variety of settings including reservoir stimulation (e.g. Karrenbach et al.,

2019; Baird et al., 2020; Dinder et al, 2020), geothermal seismicity (Li and Zhan,
2018), glacier icequakes (Walter et al., 2020; Hudson et al., 2021) and urban
monitoring (Dou et al., 2017; Song et al. 2021; Zhu et al., 2021). Earthquake
studies have used events at regional ranges (Lindsey et al., 2017; Wang et al.,
2018; Jousset et al., 2018; Ajo-Franklin et al., 2019; Sladen et al., 2019; van den
Ende and Ampuero, 2021; Nayak et al., 2021) and out to teleseismic distances
(Lindsey et al., 2020; Paitz et al., 2020). DAS recording has also been used in
the analysis of ambient noise (Dou et al., 2017; Zeng et al., 2017; Martin, 2018;
Ajo-Franklin et al., 2019; Spica et al., 2020a,b; Yang et al, 2022; Mirzanejad et al.,
2024, Smolinski et al., 2024).
Some applications of DAS systems have deployed their own fibre optic cables
(e.g., Wang et al. 2018, Walter et al. 2020), whilst many have exploited unused
fibre channels on existing cables, termed ‘dark’ fibres (e.g., Ajo-Franklin et al.,
2019). For specially deployed cable the configuration is well controlled and
shallow burial can be expected to give good ground coupling. For existing cables
although the general configuration may be known, cable loops in inspection
pits can complicate the association of DAS channels with physical position.
Also, for cables in a conduit the precise mode of deployment and the nature
of the coupling to the surrounding environment can be difficult to determine.
In urban environments the ground coupling conditions can change quite
rapidly along the cable with some modification of the DAS signal with position.
The use of existing cables has enabled studies in the marine environment,
with applications to ship tracking, the study of microseisms and recording of
earthquakes at regional and greater distances (e.g., Lindsey et al., 2019; Sladen
et al., 2019; Matsumoto et al., 2021; Rivet et al., 2021)
The strain-rate signals recorded on DAS systems have many similarities to
conventional records but differ via their directional properties and lack of
broadside response. Many authors have sought to convert DAS strain-rate
signals into equivalent ground velocity so that, for example, comparisons can
be made with co-located seismometers (e.g., Daley et al., 2016; Egorov et al.,
2018; Wang et al., 2018; Zhu et al., 2021, Lior et al., 2021). The work of Lior
et al. (2021) suggests that a minimum of 20 gauge lengths of consistent cable
will be needed to extract a reasonably accurate estimate of the apparent wave
velocity along the cable that is needed for the strain-rate conversion. However,
the geometries of many cable layouts are not well suited to such processing,
with only short lengths in a consistent direction. Many ‘dark’ fibres, particularly
in urban environments, have complex geometry linked to the configurations of
streets with only short stretches of uniform orientation so that the DAS signal
can display relatively rapid variation in amplitude. With a segmented cable
geometry, especially with a grid street pattern, strain transformation to ground
velocity is less useful. It is therefore generally more effective directly with
1.2 Structure of the work 3

the DAS response to understand the nature of the recorded seismic wavefield,
which depends on both the detailed configuration of the DAS sampling and
the orientation of the fibre relative to the components of the incoming seismic
waves.

1.2 Structure of the work

The aim of this short book is to provide and introduction to the nature of
the seismic wavefield as recorded with Distributed Acoustic Sensing (DAS) on
horizontal optical fibres in the near surface. I build on the results obtained by
many previous researchers to try to provide a unified treatment of the different
aspects of the wavefield specific to DAS recording.
Understanding of the nature of the seismic wavefield as recorded on a
DAS cable requires recognition of the effect of local spatial averaging from
the intrinsic gauge length and the important role of the orientation of cable
segments relative to the seismic waves arriving at the site. Analysis with the
aid of the frequency–slowness domain provides a convenient framework for
including such effects.
In Chapter 2 I provide a brief account of optical fibre sensing using the
interaction of laser pulses with imperfections in the fibre, with concentration on
the use of Rayleigh scattering as in DAS. The use of linked laser pulses imposes
a natural gauge length. Measurements of the change in optical path length
over the gauge around a recording channel are typically employed to extract a
measure of averaged strain-rate.
Chapter 3 summarises the way in which a representation of the seismic
wavefield can be built in the frequency – slowness domain in terms of
the reflection and transmission of seismic waves. This representation is
then employed to explain the modifications introduced by working with
gauge-length averaged strain along a cable in DAS compared with the vector
motion on a 3-component seismometer.
The dense recording provided by DAS means that the evolution of the
wavefield can be followed directly rather than inferred from patchy sampling. A
consequence is much better rendering of surface waves and other guided waves
than is achievable without very large numbers of seismometers. Chapter 4 links
the dispersion of such guided waves to the structural representation and shows
how multi-mode dispersion can provide strong constraints on structure with
depth.
The next group of Chapters (5–7) examine the character of the DAS wavefield
at local, regional and global scales. In each case examples of DAS recordings
are presented, together with a discussion of suitable ways of simulating the DAS
response for the particular regime. There is an emphasis on the specific nature
of DAS records and the way in which sensitivity to cable orientation relative to
passing wavefronts can be utilised to advantage.
4 Introduction

In Chapter 8 I turn attention to the exploitation of ambient noise via

cross-correlation of DAS channels. A simple visual approach is used to guide
understanding of the influences of cable geometry on the extraction of Rayleigh
and Love wave signals.
Chapter 9 is concerned with the response of a DAS cable configuration when
used as an array for stacking. The angular and slowness dependence of DAS
recording means that the stacked response of a DAS array is determined by
the direction and epicentral distance of the event being imaged, rather than
described by a general array function as in the case of seismometers. With a DAS
cable slower travelling wave component such as surface waves are enhanced
compared to body waves with steeper paths through the near-surface zone.
Array configurations suitable for local correlation analysis are also discussed
I hope that the various examples will encourage researchers to look at DAS
records in more detail and so understand the high fidelity rendering of the
wavefield afforded by the dense sample along a DAS cable.
2
Optical fibre sensing

A fibre optic cable used for data transmission contains a bundle of thin optic
fibres. Each fibre has a very thin core of dielectric material (about 10 µm for
single mode transmission) surrounded by a cladding with lower refractive index
than the core. The fibre is surrounded by a protective coating. The fibre acts
as an optical waveguide with the light trapped in the core with total internal
reflection at the core-cladding interface. The typical refractive index for a cable
core is 1.4475 so that the propagation speed of light along the fibre is about 69%
of the speed of light in a vacuum. The light source used for optical fibre sensor
is commonly a laser operating at near-infrared wavelengths (∼1550 nm) with
signals built up from multiple pulse transmission.
A special purpose optical cable may consist of just a few single-mode
fibres with a protective sheaf to reduce mechanical damage in handling. A
commercial telecommunications cable is likely to have upto a hundred fibres
bundled together with in a strong protective coating. Cables used for marine
installation are designed to be particularly robust.
The various optical fibre sensors depend on the interaction of the laser
light with imperfections in the fibre so that light is scattered back along the
fibre. The time difference between the initial pulse and the back-scattered
energy determines the spatial position along the fibre. Changes in the
local environment modify the character of the backscattered signal and
the differences are exploited to extract measure such as strain-rate and
temperature. A wide range of sensor systems based on optical fibres have been
developed (Udd & Spillman, 2024).

2.1 Interaction of laser pulses with fibre

Optical fibres are constructed from doped silica glass to achieve a constant
refractive index, with very tight tolerances, but very minor density variations
occur in the fibre during manufacture and as a result of handling. These
variations give rise to scattering of the light passing along the cable and
the back-scattered component can be interrogated at an instrument at the
transmission site. The configuration of the density variations and consequent

5
6 Optical fibre sensing

Figure 2.1 Schematics of the character of the Rayleigh, Raman, and Brillouin peaks in
the back-scattered light spectrum. T represents temperature and ² is strain

scattering is not actually known, but the distribution should be time-invariant

and is assumed to be dense and relatively homogeneous. As noted by Lindsey
et al. (2020) the common telecommunications loss standard for optical fibres of
0.15 dB/km for a wavelength of 1,550 nm means that a few thousand photons
are lost from the transmitted light along the fibre every metre.
A change in the environment of the fibre can cause the local length, diameter
and refractive index to change, and in consequence modify the character of the
back-scattered light. Changes in the amplitude, frequency, and phase of this
back scattering can be linked to local variations in temperature and strain out
to distances of tens of kilometres or more.
The local scattering is of three types: Rayleigh scattering with almost no
change of frequency, and Brillouin and Raman scattering with the possibility of
both frequency decrease (Stokes) and increase (anti-Stokes). The Rayleigh band
has the highest intensity of scattering, followed by Brillouin scattering and the
weakest effect is in the Raman bands.
Rayleigh scattering of electromagnetic waves in the fibre is associated with
local variations with sizes well below a nanometre and does not involve energy
transfer to the scatterer. As a result, the frequency of the scattered wave is almost
the same as the incoming, a Doppler shift is possible in the case of significant
movement of the scatterer. The intensity of scattering varies with the fourth
power of the optical frequency.
Both Brillouin and Raman scattering of electromagnetic waves are inelastic
processes involving energy transfer to the scatter. Where energy is absorbed by
the scatterer the scattered wave has lower frequency, but if energy is lost from
the scattering zone the frequency of the scattered wave is raised.
2.2 DAS exploiting Rayleigh scattering 7

Interrogator Unit

Laser P2 P1 L-g/2 L+g/2 Optical fibre

Detector & Gauge
Interferometer S2 S1 Length (g)

Figure 2.2 A DAS interrogator sends laser pulses P 1 , P 2 separated by a gauge length
g along an optical fibre, the evolution of the relative phase of the Rayleigh scattered
returns S 1 , S 2 from around a channel location L determines the changes in the optical
path length along the gauge length and thence strain rate. The spacing of sampling
channels can be smaller than the gauge length, but then the results from successive
channels are not independent.

Brillouin scattering involves interaction with the local lattice structure of the
silica based fibre, with a frequency shift that depends on the acoustic wavespeed
and thus shows a linear dependence on temperature and strain. The minimum
frequency shift is about 11 GHz at 1530 nm.
Raman scattering involves much larger frequency shifts (∼ 13.0 THz at 1550
nm) associated with modifying the local vibrational state of the scatterer. The
Stokes component is almost independent of temperature whilst the anti-Stokes
component with higher frequency is temperature dependent. Although Raman
scattering is weak it provides a way to sense the temperature distribution along
a fibre.
2.2 DAS exploiting Rayleigh scattering
Distributed Acoustic Sensing (DAS) exploits the Rayleigh back scattering from
laser signals sent into an optical fibre using optical time domain reflectometry
(OTDR). A coherent laser pulse is injected into the fibre and the net result of
the interference of the back-scattered waves is returned to the excitation point.
With local deformation of the fibre an optical phase shift is introduced between
back-scattered components from the leading and trailing parts of the optical
pulse passing through the deformed zone. With modified interference the
intensity of the returned light changes from pulse to pulse. Both the intensity
and phase variations can be exploited in different classes of DAS system.
The randomness of the density fluctuations within the fibre means that the
returned intensity varies randomly with time lag and hence distance along
the fibre, but the phase varies more smoothly with time. DAS systems rely
on measuring the phase difference between the backscattered light from two
separate returned pulses by an interferometric process (Figure 2.2). The pulses
are separated in time, often with an optical loop, and the time delay imposes a
"gauge length" over which variations in phase are measured. This gauge length
is half the distance traversed between the passage of the two pulses. The effect
of the time delay allows the DAS system to produce phase results for a a suite
of channels along a fibre, these are not point measurements but span the gauge
length about the channel location.
8 Optical fibre sensing

The differences in phase between the backscattered pulses are proportional

to the change in optical path length over the gauge length associated with local
strain. A further challenge is then to extract the phase differences in a way that
provides a broad dynamic range. This process requires unwrapping the phase
to a continuous function, rather than just confined to the interval −π to π. By
analysis of multiple sequences of sampling pulses in time the resultant output
is in terms of strain rate.
A variety of different approaches to phase extraction are employed, with
efforts made to reduce instrument induced noise. Current DAS systems can
capture dynamic strains over a very broad range of frequencies 0.001 – 1,000
Hz (e.g., Paitz et al., 2020), and so can be used in a wide range of seismological
contexts, even though they only provide uni-directional information.

2.2.1 Relation of DAS signal to strain rate

The phase difference between returned pulses separated by a gauge length g is
controlled by the change in optical path length. For a segment about a specific
channel at position L the time difference between the returned pulses is
Z L+g /2 1 2
Z L+g /2
τ(L, t ) = 2 dx = dx n(x, t ). (2.2.1)
L−g /2 v(x, t ) c 0 L−g /2

where v is the speed of light in the fibre segment, c 0 the speed of light in vacuo
and n the refractive index.
In the presence of a seismic signal there will be change in the time difference
arising from the combination of the change in the length of the gauge segment
and the modification of the speed of light in the fibre in the presence of strain.
As in Kennett & Bunge (2008) we use coordinates x to represent the current
state and ξ the undeformed state. For a fibre element under deformation, the
current length dx is related to the undeformed length dξ by the stretch λk , so
that

dx = λk dξ (2.2.2)

The axial Green strain e k along the fibre is given by

e k = 12 (λ2k − 1) and so λk = 1 + 2e k
p
(2.2.3)

With an expansion in terms of powers of e k the differential stretch λk can be

represented by

λk = 1 + e k − 34 e k2 + 15 3
8 e k + .... (2.2.4)

Except in the immediate vicinity of a source, the quadratic and higher terms will
be insignificant.
Working in terms of the undeformed state, the rate of change of the time
2.2 DAS exploiting Rayleigh scattering 9

difference is thus
∂ 2 ∂ L+g /2
Z
τ(L, t ) = dξ λk n(x(ξ), t )
∂t c 0 ∂t L−g /2
2n 0 ∂ L+g /2 δn(x(ξ), t )
Z · ¸
= dξ λk 1 + (2.2.5)
c 0 ∂t L−g /2 n0
where n 0 is the refractive index in the undeformed segment and δn arises as a
consequence of deformation.
The modification of the speed of light in the fibre, and thus the refractive
index, due to deformation arises from the photoelastic effect that depends on
both the axial strain e k and the transverse strain e ⊥ (see, e.g., Kuvshinov, 2016).
For a typical optical fibre
δn
= −[0.3e k − 0.2e ⊥ ]. (2.2.6)
n0
Combining the direct strain effects and the those due to the modification of
the light speed we have
∂ 2n 0 ∂ L+g /2
Z
τ(L, t ) =
£ ¤
dξ [1 + e k ] 1 − [0.3e k − 0.2e ⊥ ] . (2.2.7)
∂t c 0 ∂t L−g /2
Working to first order in strain we extract
∂ 2n 0 L+g /2 ∂£
Z
τ(L, t ) =
¤
dξ 0.7e k − 0.2e ⊥ . (2.2.8)
∂t c 0 L−g /2 ∂t
For weak perpendicular strain ²⊥ the Poisson effect enhances the apparent
longitudinal strain. For an optical fibre with a typical Poissons’s ratio of 0.2, the
DAS signal
∂ 2n 0 ζ L+g /2 ∂e k
Z
τ(L, t ) ≈ dξ , (2.2.9)
∂t c 0 L−g /2 ∂t
where the effective refractive index n 0 ζ ≈ 0.8n 0 .
The transversal strain term in (2.2.8) only becomes significant when the axial
strain is small, such as when waves arrive broadside to the cable very close to
a source. Nevertheless the contrasts in material properties between the cable
packing and the stiff optical fibres acts to reduce the effective transverse strain
of the fibre compared with that acting at the exterior of the cable (Budiansky et
al., 1979; Kuvshinov, 2016). In favourable circumstances, the transversal fibre
strain may reach about 5% of the transverse strain on the cable itself, but will
generally be less.
Since strain-rate DAS measurements are dominated by the axial strain along
the optical fibre, they represent just a single component of the strain tensor.
This means that the effect of a change in the direction of incoming waves by
an angle θ produces tensorial strain rotation that depends on functions of 2θ.
Whereas for the horizontal components of seismometers the vector rotation
10 Optical fibre sensing

depends on functions of θ (e.g. Martin, 2018; Zhan, 2020). As a result, the

orientation of the optical fibre relative to incoming seismic waves plays an
important role in determining the relative amplitude of DAS signals. These
effects are important for both local and more distant sources.
A variety of different hardware implementations have been proposed for DAS
interrogation. These include the recording of a “deformation-rate” quantity ṽ
that under ideal conditions would be equivalent to the component of ground
velocity resolved along the axial direction of the fibre (e.g., J Yang et al., 2022).
Again two pulses at a fixed time interval are employed and correlation is made
between the returned scattered primary pulse and the optically delayed second
pulse, to produce the integral of strain rate ė from the interrogator to the desired
channel (x)
Z x
ṽ(x, t ) = ds ė(s, t ) (2.2.10)
0

For this quantity the directional sensitivity is equivalent to a seismometer.

2.2.2 Processing issues
The Interrogator unit sends laser pulses down the fibre at regular intervals.
In the simplest configuration the interval between pulses is set so that light
scattered at the far end of the cable is received before the next pulse is sent.
A longer pulse width injects more light into the fibre and so improves signal
to noise, but should be shorter than half the time of passage across the gauge
length so that there is no overlap between returns from the leading and trailing
edge of the gauge.

Gauge length
The choice of gauge length g plays an important role in the nature of the DAS
response, since it needs to be long enough to capture a measurable phase delay,
yet not so long as to impose signal distortion. Longer gauge length suppresses
the effects of random noise and so acts to improve signal to noise, but has to be
balanced against signal quality.
The effect of gauge length can be described as a moving-average filter with
respect to position along the fibre (Dean et al., 2017). This averaging imposes
notches in the response for those frequencies that are multiples of the inverse
of the passage time of a wave across the gauge length (Dean et al., 2017;
Karrenbach et al., 2019). This suite of frequencies are multiples of the apparent
wave velocity along the cable divided by the gauge length (see §3.4). Gauge
length averaging also means that for low frequency waves with long wavelengths
there is little variation across commonly used gauge intervals, and so they tend
to be poorly recorded by DAS (e.g., Spica et al., 2020).
A further consequence of gauge length averaging is that a localised strong
effect, such as that due to the immediate vicinity of a source, will affect a span
of channels over a distance comparable to the gauge length.
2.2 DAS exploiting Rayleigh scattering 11

Fading
The backscattered signal returned from the distribution of density fluctuations
in the fibre represents the superposition of many scattered waves whose net
effect can sometimes be close to zero intensity. In such circumstances the
estimation of phase becomes very noisy. The effect of such fading changes
in both distance and time, for example very minor temperature changes can
shift scattering sites and cause temporary reduced intensity. At any time a few
percent of channels may exhibit fading and so channel stacking procedures are
built into the DAS interrogator system to enhance signal to noise. A simple
approach is to stack over a few channels around the nominal channel position.

2.2.3 Cable coupling

A DAS system renders the strain or strain rate along the optical fibre being
sampled, but how far this represents the situation in the surrounding materials
depends on the local coupling conditions for the cable containing the fibre (e.g.,
Ajo-Franklin et al., 2019). In a borehole, a free hanging cable has no natural
clamping to the well wall. However, for a permanent installation the cable can
be cemented behind the casing with direct coupling to rock.
Where the installation of a cable is under user control, e.g. in a trench,
coupling to the surroundings can be expected to be strong, though may
vary depending on local conditions. For horizontal near-surface cables in
conduits, such as dark fibre on existing telecommunication cables, the effect
of gravity and friction is generally sufficient to provide adequate linkage to the
environment, though conditions can vary significantly along a cable run.

2.2.4 Channel calibration

The position of a channel relative to the fibre can be determined from the
passage time for backscattering with knowledge of the light speed for the fibre,
and then correlated with information on the configuration of the installation.
However, for cable in conduit, the actual geometrical configuration will depend
on the nature of the conduit since extra cable length may be allowed to avoid
cable tension. Further, cable loops may be introduced at inspection points to
take up excess length.
This means that specific calibration is needed to constrain the actual
geometry, commonly using “tap tests” at controlled points with independent
location information (e.g., with GNSS observations). With a concentrated
source, the strongest amplitude and earliest time of arrival in the DAS response
indicates the channel location. However, the nature of the DAS processing
means that it is difficult to achieve spatial localisation to to better than about
half the gauge length, even when all channels are available (e.g., Kennett et al.,
2024).
3
Representing the seismic wavefield for DAS

In this chapter we exploit the representation of the seismic wavefield in terms of

a superposition of frequency and slowness components to provide a description
of the seismic wavefield for stratified media and show the way in which this can
be exploited to produce practical results for DAS response.

3.1 Strain rate and ground velocity

As we have seen in the previous Chapter, DAS recording relies on exploiting the
changes in optical path length over a gauge length around a nominal recording
point. The relative change in optical path length along this gauge length due
to disturbances passing across the fibre is dominantly controlled by the axial
strain rate e d within the gauge. This strain rate is just the spatial derivative of
the ground velocity along the fibre v d with respect to distance along the fibre.
Under the assumption of uniform sampling along the gauge length g , the
effect of averaging the axial strain rate around the reference point takes the form
1 g /2 ∂v d (s) 1 h 1h
Z ig /2 i
〈ė d 〉 = ds = v d (s) = v d (g /2) − v d (−g /2) (3.1.1)
g −g /2 ∂s g s=−g /2 g
The averaged strain rate can therefore, in principle, be obtained by differencing
the ground velocity resolved along the cable at the ends of the gauge length.
This connection was first noted by Bakku (2015) using a plane wave model
for vertical seismic profiling, and has been widely employed for modelling DAS
response for close sources in borehole configurations (e.g., Daley et al., 2016;
Binder et al., 2020).
When an event is close to the DAS cable, the change in the waveforms across
the gauge length can be significant due to the strong geometrical variations
(Figure 1) and so the appearance of the DAS wavefield does not directly
correspond to what is seen in ground velocity on a geophone (Parker et al.,
2018).
Since the DAS response to a seismic disturbance can be directly related
to knowledge of the corresponding ground velocities at the end of the gauge
length we are able to make use of existing methods of treating the seismic

12
3.1 Strain rate and ground velocity 13

sc-g/2 ψ1 sc sc+g/2 ψ2 cable

v1 v2
x1 x2
d0
source

Figure 3.1 Plan view of a near source to a DAS cable. For a nominal cable channel at
s c and a source with closest distance to the cable of d 0 , the velocity resolved along the
cable v 1 , v 2 has to be calculated for the points s c ± g /2, which lie at different distances
x 1 , x 2 from the source and with different inclinations of the radial vector to the cable
ψ1 , ψ2 .

wavefield. We are therefore able to build on representations of the response

of a stratified medium to a source with a expansion in terms of plane waves in
the frequency–slowness domain (e.g., Kennett, 2001).
Consider a suite of DAS channels at locations {xc }, and describe the cable
configuration around the channel point via the local unit tangent vector d(xc ),
under the approximation of local linearity along the gauge length. This
assumption generally works well except in the vicinity of a sharp turn in fibre
direction (see §3.6). For a position xc the averaged strain rate along the cable
direction d(xc ) can be expressed as
Z Z
〈ė d 〉(xc , ω) = d2 p 〈ė d 〉(p, ω)eiω[Link] , (3.1.2)

where p is the local slowness vector in the horizontal plane, with magnitude
p. In terms of ground velocity differences (3.1.1) we have the alternative
representation
1
Z Z
〈ė d 〉(xc , ω) = d2 p [v.d](p, ω) eiω[Link]+ − eiω[Link]− ,
£ ¤
(3.1.3)
g
where xc± = xc ± g d/2. The component [v.d](p, ω) is in common for the two
ends of the gauge length and can be expressed in terms of the reflection and
transmission properties of the medium. With the introduction of a source at
the origin, the double integration over horizontal vector slowness p can be
recast as an integral over slowness p accompanied by an angular expansion
in vector harmonics (e.g., Kennett, 2001) as in the usual approach to synthetic
seismograms for a layered medium. The differencing associated with the gauge
length means that it is necessary to take account of the differences in range and
azimuth relative to the DAS cable at the two ends. As illustrated in Figure 1,
for sources close to the cable line there can be noticeable differences in these
quantities.
14 Representing the seismic wavefield for DAS

For more distant sources the distances and angle subtended at the two ends
of the gauge length will be very close and so subtraction of ground velocity
contributions will be subject significant loss of numerical precision. It is
therefore more effective to work with the composite response (see §3.4.2).

3.2 Building the wavefield

Working in the frequency – slowness domain we can exploit the results
developed in Kennett (1983, 2001) for the character of the wavefield in terms
of the reflection and transmission properties of portions of a stratified model
excited by upgoing and down going waves from the source.
The first stage of generating the representation of the wavefield is to include
the description of the radiation from a seismic source. For the initially
downward radiated energy, reflection from below the source returns energy
to the source level and this is then subject to multiples generated at the free
surface. The final stage is to include transmission to the surface and the
influence of near-receiver structure. Double integration over slowness and
frequency can then be employed to extract synthetic records for specified
receiver points

3.2.1 Structural response

We consider a receiver placed at the surface and a structural model occupying
the region between the free surface and depth z L . Beneath L the velocity
structure is assumed to be uniform so that a radiation condition can be
applied to allow only purely downgoing waves propagating into this region. A
source is introduced beneath the surface at the depth z s , The situation with a
surface source can then be described by taking the limit as the source depth
tends to zero. The source is envisaged as lying in a locally uniform medium,
which means it is possible to define distinct upward [U U S ] and downward [DDS]
radiation. The source is introduced in terms of the jumps in displacement and
traction across the source plane and this means that moment tensor or single
force systems can be readily described.
We now regard the region above the source to abut the uniform medium and
to be irradiated by the upgoing field U S . Similarly the region below the source
is taken to have an incident downward field D S from the uniform medium. This
allows us to represent the development of the seismic wavefield in terms of
sequence of propagation processes each of which can be described by reflection
and transmission operators for the regions f S and SL, above and below the
source, as illustrated in figure 3.2.
We start with the energy radiated upward from the source. This produces a
fS fS
direct wave at a surface receiver described by WU U S where WU is the operator
which generates the surface response from an upgoing field at the level S. The
upgoing wave U S will be also be reflected back by the region above the source
3.2 Building the wavefield 15

fS SL S fS S fS S
WU RD [D + RU U ] WU [U ]
f
fS fS
RU RU S
U
S
D
S *
S fS S
[D + RU U ]
SL SL
RD RD

Figure 3.2 Schematic presentation of the development of the seismic wavefield excited
by a source at depth z s , in a stratified halfspace, in terms of the action of reflection and
transmission operators for the different parts of the structure.

(particularly from the free surface itself ) and this will give a downgoing wave
fS
contribution at S of RU U S . The total downgoing wave field at S is therefore
fS
composed of the source radiation D S and the surface “ghost” RU U S . This
“ghost" includes the depth phases such as pP, sP, sS and pS, with conversion
occurring in the reflection process at the surface.
fS
The combined downward field D S + RU U S is now incident on the lower part
of the model SL. With a single reflection this produces the upgoing contribution
fS
RSL
D
D S +RU U S ] at the source level. The reflection operator RSL
[D D
will include the
full effect of the structure beneath the source, including internal multiples. With
fS fS
transmission to the surface a further contribution WU RSL D
D S + RU U S ] will be
[D
added to the surface response. In addition there will be a further reflection
from the zone above the source including the free surface that again will be
fS
reflected back from below the source level acquiring the combination RSL D U
R
before transmission to the surface. Each successive reflection above and below
fS
the source level requires a further application of RSL R .
D U
The seismic disturbance w o at the surface, is therefore composed of a
propagation sequence of the form

f S −1 SL
³ ´
fS fS fS
w o = WU U S + WU I − RSL
D RU D S + RU U S ].
RD [D (3.2.1)

fS
The inverse operator (I − RSL
D U
R )−1 includes all multiple interactions between
the regions above and below the source:

f S −1
³ ´
fS fS fS
I − RSL
D RU = I + RSL SL SL
D RU + RD RU RD RU + ...... (3.2.2)

In the representation (3.2.1) for the wavefield the reverberation operator for
16 Representing the seismic wavefield for DAS

the entire structure is combined with reflection from below the source level.
Since we have the operator identity
f S −1 SL f S SL −1
³ ´ ³ ´
I − RSL
D RU RD = RSL
D I − RU RD , (3.2.3)

we can view this reverberation sequence as occurring on either the source or

receiver sides of the main reflection RSL
D
.
For a surface source, we can envisage the level S brought right up to the
fS
surface so that D S + RU U S will just be the net downward radiation from the
source.

3.2.2 Constructing synthetic records

For a stratified model, the reflection and transmission operators take the form
of matrices in the frequency–slowness domain, and synthetic seismograms
can be constructed by double integration over slowness and frequency with
appropriate phase factors for the distance from source to the receiver point.
For cylindrical symmetry around the source point this means an expansion in
vector harmonics, with integration over radial slowness.
A typical surface DAS cable lies at shallow depth, from just beneath the
surface to 2 m, or so, depending on the particular configuration. This means
that there is a slight time offset between up-going waves arriving at the fibre
and the down-going waves reflected at the free surface. The delay is small,
and can normally be neglected unless the surface wavespeeds are very low or
very high frequencies are used with active sources. The interference effects
from shallow cable emplacement can be included by replacing the free-surface
reflection coefficients that depend only on slowness by a frequency dependent
term representing the interaction of up-going and down-going waves at the
cable depth.
For a stratified model overlying a halfspace with P wavespeed αL and S
wavespeed βL , there are singularities at the corresponding slownesses α−1 L , βL
−1

that give rise to the need for branch cuts in the complex slowness plane to satisfy
the downgoing wave condition in the lower halfspace. For slownesses greater
than β−1
L a set of poles arise when determinant of the inverse reverberation term
in (3.2.1 vanishes, i.e., when
³ ´
fS
det I − RSL
D RU = 0. (3.2.4)

In the presence of seismic attenuation these poles move off the real slowness
axis, into the upper half plane when ω > 0 for frequency dependence exp(−iωt ).
For P-SV waves the pole for the fundamental Rayleigh mode can occur beyond
the surface S wave slowness β−10 with a limit at p R0 – the Rayleigh wave slowness
for a halfspace with the properties just at the surface. All higher Rayleigh modes
are limited to slownesses less than β−10 . For SH waves the limit is β0 for the
−1

fundamental and higher mode Love modes.

3.2 Building the wavefield 17

Im p 1
0 0
1
0
1 0
1 D
0
1
0
1 0
1
0
1
0
1 0
1
0
1
0
1 0
1
0
1
0
1
0
1 0
1
0
1
Branch cuts
0
1 0
1
0
1 0
1
0
1 0
1
0
1 0
1
0
1 0
1
0
1
0
1 0
1
0
1
0
1
0
1 0
1
0
1 Poles Re p
0
1 0
1
0
1 0
1 x x x x x x x
α -1
L
β -1
L
β 0-1 pR0 C

p plane

Figure 3.3 The singularities, branch cuts and integration contours in the p plane for
the full response of a stratified half space, with underlying P wavespeed αL and S
wavespeed βL , for fixed frequency ω.

The most direct approach to the evaluation of the integration over slowness is
to perform a direct numerical integration along the real p axis. This works well
for an attenuative medium, though the presence of nearby poles can require
fine sampling in slowness particularly for high frequencies. Such calculations
include both body and surface waves.
Attention can be focussed on the contributions from the poles representing
surface wave contributions by deforming the contour C in Figure 3.3 into the
upper half plane to D. This picks up all the polar residue contributions for
slownesses greater than β−1L .
The distribution of poles depends strongly on frequency. For low frequencies
only a few poles arise, whilst at higher frequencies there are a great many.
At even moderate frequencies the number of modal residue contributions
N (ω) becomes very large indeed, and locating all the poles is significant
computational problem.
The poles with largest slowness give the major contribution to what would
generally be regarded as the surface wave train, with relatively low group
velocities. The summation of modes with smaller slownesses provides a means
of synthesising S body wave phases by modal interference. A residue sum taken
over a portion of the real p axis can provide good results for the S wave coda
(Kerry 1981).
The pattern of singularities will be modified when approximations to the
response are made by singling out specific features from the general response
(3.2.1) but direct numerical integration in slowness for an attenuative medium
will usually give good results.
18 Representing the seismic wavefield for DAS

3.3 DAS - strain rate and orientation

An important component for evaluating the response of a DAS cable to seismic
waves from a source is a good knowledge of the geometrical configuration of the
channels, since this defining the issues arising from the orientation of the cable
relative to the incoming wavefield.
For a stratified medium, DAS synthesis can be in the frequency-slowness
domain using the formulation above in terms of the reflection and transmission
properties of layers coupled to suitable source forms. For close sources a direct
differencing approach can be employed, whereas for more distant sources an
analytical representation of the differencing operator is more effective.

3.3.1 Local field

For sources close to a optical cable the representation (3.1.1) can be used
directly in modelling by differencing synthetic seismograms calculated for the
potions at the two ends of the gauge interval. A direct test of the relation
(3.1.1) has been made by Wang et al. (2018) by constructing the difference of the
output of two seismometers spanning the interval around a set of DAS channel
locations for the frequency band 1–5 Hz at the Brady test site. Even though
the distance between the seismometers spanned a number of channels, the
summed DAS response over these channels was in good agreement with the
difference between the ground velocity from the seismometers resolved along
the cable direction. The configuration was equivalent to a rather long gauge
interval.
For close events the representation (3.1.1) can be used directly in modelling
by differencing synthetic seismograms constructed from 1-D velocity profiles.
For example, Hudson et al. (2020) use reflectivity style calculations of Green’s
functions for inversion of the mechanism of microseisms at the base of an ice
stream in Antarctica. Wang et al. (2018) show a few examples of calculated traces
at regional distances produced by a similar approach.
As noted above, relation (3.1.1) means that, in principle, we can work
directly with existing ways of calculating synthetic seismograms, but need to
include the differencing effects of the gauge length. This can be done by
explicit differencing of prior ground velocity calculations provided that the
geometrical configuration is fully specified (e.g. Wang et al., 2018; Hudson et al,
2020). The alternative is differencing whilst constructing the synthetics in the
slowness-frequency domain as in (3.1.3), which has the advantage that many
channels can be calculated exploiting the same medium response terms once
the cable configuration has been mapped. Direct differencing is most effective
for close sources to the DAS fibre.

3.3.2 Distant field

When the DAS cable lies well away from the source of seismic waves the
3-component ground velocity field has similar character across even a kilometre
3.3 DAS - strain rate and orientation 19

or two of cable. The main effects now come from the orientation of the cable
with respect to the incident arrivals. In such a situation the arriving wavefronts
are approximately plane, the differences in event distance at the ends of the
gauge interval are very small, and the ground velocities and orientations are
very similar. As a result, there is a strong likelihood of significant loss of
numerical precision when the velocity traces are differenced to produce strain
rate using (3.1.1). It is preferable for regional and teleseismic distances to work
with analytic representations of the differencing within a slowness-frequency
integration that can readily capture orientation effects along a complex cable
geometry.
With frequency-slowness synthesis, we build the response from a spectrum
of plane waves, and can develop a formalism for the effect of gauge length on
distant events. The formulation here for a horizontal DAS cable builds on the
analysis in the supplementary material to Walter et al. (2020).
Consider a plane wave with frequency ω that has propagated a distance ∆
to a reference point xc from a distant source with horizontal slowness vector p.
In the neighbourhood of this reference point, the displacement associated with
the plane wave is

u = u neiω[p.x+p∆] , (3.3.1)

where x is the offset from the reference point and n is the polarization vector for
the plane wave. The associated strain tensor
1 ∂u i ∂u j
µ ¶
iω
ei j = + = u (n i p j + n j p i )eiω[p.x+p∆] . (3.3.2)
2 ∂x j ∂x i 2
The strain resolved along a DAS cable lying in the local tangent direction d at xc
is then

e d = d i e i j d j = iωu (d.n)(d.p)eiω[p.x+p∆] . (3.3.3)

The DAS measurement associated with a point represents an average of strain

along the gauge length g along the cable around that point. Thus, setting x =
sd aligned along the local cable direction, the averaged strain rate around the
reference point
1 g /2 iω
Z h ig /2
〈ė d (ω)〉 = ds ė d (s) = − u (d.n) eiω[sd.p+p∆]
g −g /2 g s=−g /2
2ω ³ ωg ´
= u (d.n) sin d.p eiωp∆ . (3.3.4)
g 2
The term d.n depends on the polarization of the relevant wave type. The
combination p d = d.p represents the slowness of the plane wave resolved along
the cable, and so p d = p cos ψ as a function of the inclination of the cable ψ
relative to the radial direction from the source.
Relations equivalent to (3.3.4) have been derived in many ways with differing
20 Representing the seismic wavefield for DAS

notation. Many forms employ the effective propagation velocity along the cable
c d = 1/p d = 1/(p cos ψ) rather than slowness p. For high frequency waves the
full form of (3.3.4) is needed, and has been applied by Egorov et al. (2018) to
convert averaged strain to ground velocity for vertical seismic profiling so that
they can employ standard methods for full waveform inversion. The result can
be simplified further for lower frequencies.
When the variation of the plane wave along the cable is slow compared to the
gauge length, i.e., ωg p d /2 ¿ 1, the sine in (3.3.4) can be approximated by its
argument. With this simplification,

〈ė d (ω)〉 = ω2 u (d.n)(d.p)eiωp∆ . (3.3.5)

Thus the averaged strain rate is equivalent to the ground acceleration due to the
plane wave resolved along the cable modulated by the slowness along the cable.
Under this approximation, the averaged strain rate 〈ė d 〉, in the general
case, can thus be constructed by combining the radial u r and tangential u t
components to give

〈ė d (p, ω)〉 = ω2 p cos ψ(u r (p, ω) cos ψ + u t (p, ω) sin ψ)eiωp∆ . (3.3.6)

Although the slowness factor is universal, the radial and transverse components
will also be functions of slowness, and their properties will depend on the
polarisation of the waves. This representation of the effect of gauge length will
be suitable for frequencies

f < f c = 1/(5πg p), (3.3.7)

since the strongest variation occurs when the cable and propagation path are
aligned. For regional phases with a gauge length of 10 m, f c ∼ 40 Hz for P waves,
and f c ∼ 20 Hz for S waves. The values of f c are even higher for teleseismic
waves, with a steeper inclination to the vertical and so smaller horizontal
slowness.
The approximate result (3.3.6) does not depend on the precise nature of the
averaging along the gauge length. The relation between averaged strain and
ground velocity v d projected along the cable is often given in the apparently
simpler time domain form

〈e d 〉 = v d /c d , (3.3.8)

(e.g., Daley et al., 2016) and forms the basis of conversion schemes to extract
ground velocity from DAS records for sources lying well away from the cable
(e.g. Wang et al., 2018; Lindsey et al., 2020; Paitz et al., 2020).
The approximation (3.3.6) is suitable for both regional and teleseismic
distances, and can be implemented by simple modifications of the receiver
terms in the frequency-slowness response (3.2.1) before integration. To apply
(3.3.6) the source needs to be sufficiently far from the array for the local field to
3.3 DAS - strain rate and orientation 21

appear as a passing plane wave, so that for frequencies around 1 Hz a minimum

distance of a few tens of kilometres is needed for natural sources.
The DAS response term (3.3.6) applies a factor of horizontal slowness to
acceleration, irrespective of the relative orientation of the cable and the
incoming wave. This means that steeply arriving waves with small slowness,
such as P and S body waves from distant sources are diminished, whilst the late
surface waves with large slowness are amplified. As a result, Rayleigh waves are
even more pronounced on broad-band DAS records of distant events than on
comparable seismograms (Lindsey et al., 2020), and the surface wave coda is
extended. The amplification of locally scattered surface waves can also have the
effect of reducing the coherence of DAS recording of body waves (van der Ende
& Ampuero, 2020).
When the condition (3.3.7) is satisfied there is simple scaling between
projected ground velocity and averaged strain rate linked by the effective
slowness along the fibre (3.3.8). This relation has been used by a number of
authors (e.g., Wang et al., 2018; Lindsey et al, 2020) to implement conversion to
ground velocity using f −k mapping, but such conversion requires a significant
length of fibre with a common orientation, and needs careful regularization
when data has only a narrow wavenumber content. An alternative is to estimate
the apparent slowness for the portion of the records being considered, e.g., Paitz
et al. (2020) have used Rayleigh wavespeeds for the surface wave portions of
recordings in Switzerland of events in Greece and Fiji.
As noted by Walter et al. (2020) the action of the DAS strain-rate averaging
in (3.3.6) is to change the amplitude spectrum of the wavefield, whilst leaving
the phase associated with propagation from the source unchanged. This means
that travel-time picking is unaffected by the DAS recording system provided the
frequency is not too high.

3.3.3 Orientation factors

The net result of the tensorial projection of strain onto the oriented cable is
that the behaviour is controlled by the relative angle ψ between the passing
wavefront and the DAS cable. This is strictly true only when the source is far
enough away for the wavefront to be essentially plane, but can be conveniently
used for even close sources to provide a useful assessment of relative amplitude
behaviour.
For a horizontal cable, the radial component of acceleration from P and SV
waves has the projection factor

u r cos2 ψ = u r 12 (cos 2ψ − 1) (3.3.9)

and for the tangential component, from SH waves,

u t sin ψ cos ψ = u t 12 sin 2ψ (3.3.10)

The radial contribution is always positive, whilst the transverse contribution

22 Representing the seismic wavefield for DAS

90 90
(a) 120 60
(b) 120 60

150 30 150 30

180 0 180 0

210 330 210 330

240 300 240 300

270 270

Figure 3.4 Comparison of seismometer and DAS strain-rate directional sensitivity in

the horizontal plane. The seismometer results for the central point are shown in light
blue and the DAS in red. The wavelength is assumed to be long compared to the gauge
length marked in black. Dashed lines indicate negative zones. (a) P waves and Rayleigh
waves; (b) S waves. [After Zhan (2020)].

can change sign. The double angle dependence (2ψ) comes from the tensor
projection of strain and contrasts with the simple vector rotation for the
horizontal components of seismometers that depends on the angle (ψ) itself.
Martin (2018) provides a detailed discussion of orientation effects for different
classes of waves, both for direct measurement and for cross-correlation as in the
analysis of ambient noise.
For Rayleigh waves, the radial orientation term cos2 ψ is to be applied, with
maximum response along the DAS cable (e.g. Martin, 2018, Zhan, 2020). For
Love waves, the tangential factor sin 2ψ comes into play and this 4-lobed pattern
has nulls for waves travelling along and perpendicular to the DAS cable (Figure
3.4). As a result, Rayleigh waves from distant events tend to be more prominent
on DAS recordings than Love (e.g., Lindsey et al., 2020).
Since DAS records are commonly plotted in terms of distance along cable
without consideration of orientation factors it is convenient to have a summary
of such effects. This can be provided by plotting out the orientation terms
on the same scale as the record section. Examples are given in Chapters 6
and 7 for regional and teleseismic events across arrays with notable changes of
geometry. Such diagrams can be helpful in disentangling effects arising from the
interference of different phases from changes in direction of the cable layout.
The simple directional dependence shown in Figure 3.4 is modified for
wavelengths of the same order as the gauge length: cf. (3.3.9) and (3.4.2). As
illustrated by Martin et al. (2021) the behaviour becomes quite complicated
when the gauge length is comparable to the wavelength, with additional lobes
in the directional response.
3.4 Gauge length issues 23

For DAS systems providing “deformation-rate” output, the orientation factors

are equivalent to those for seismometers, i.e. cos ψ for the R component and
sin ψ for the T component. As can be seen from Figure 3.4 , there is only modest
change in angular dependence for P waves compared with the use of strain rate,
but a very different behaviour for S waves. In particular, sensitivity to nearly
vertically travelling waves is improved.

3.4 Gauge length issues

The effect of the gauge length g in DAS recording is to modulate the apparent
response in frequency and slowness space (ω − p).
For a surface wave with slowness p the averaged strain rate over a gauge length
can be written as in (3.3.4),
2ω ³ ωg ´
〈ė d (ω)〉 = u (d.n) sin d.p . (3.4.1)
g 2
where u is the amplitude, d is a unit vector along the local cable direction and n
is the polarization vector for the plane wave.
For Rayleigh waves oriented along the radial direction, the slowness projected
along the cable p d = d.p = p cos ψ, and d.n = cos ψ in terms of the inclination ψ
of the cable relative to the radial direction from the source.
Thus the averaged strain rate
2ω ³ ωg ´
〈ė d (ω)〉 = u cos ψ sin p cos ψ . (3.4.2)
g 2
In terms of the phase speed c = 1/p we have
2ω ³ ωg ´
〈ė d (ω)〉 = u cos ψ sin cos ψ . (3.4.3)
g 2c
Simplifying to the in-line case, the frequency - phase speed response becomes
4π f ³πf g ´
〈ė d ( f )〉 = u sin . (3.4.4)
g c
with nulls when the frequency f = nc/g . Hence increased gauge length reduces
the frequencies at which the nulls become operative. Once the frequency is
greater than c/g the sign of the averaged strain-rate flips and switches back
again at 2c/g . These change will affect the overall phase and further corrupt
the apparent dispersion.
We can illustrate the effect of varying gauge length by plotting up the
frequency–phase speed kernel in the same way as used for representing
dispersion curves in Chapter 4. Figure 3.5 shows the behaviour for a gauge
length of 10 m. Over most of the ranges in frequency and phase speed the
modulation due to averaging is only slowly varying. However for frequencies
above 40 Hz and low phase speeds there is a rapid change around the
pronounced null.
24 Representing the seismic wavefield for DAS

100.0
Gauge length 10 m
90.0

80.0

70.0
100
60.0
Frequency [Hz]

50.0

40.0 0

30.0

20.0
-100
10.0

0.0
0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00
Phase speed [km/s]

Figure 3.5 Frequency - phase speed behaviour of the effect of gauge length averaging
in DAS recording.: Gauge length: 10 m
100.0
Gauge length 20 m
90.0

80.0

70.0
100
60.0
Frequency [Hz]

50.0

40.0 0

30.0

20.0
-100
10.0

0.0
0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00
Phase speed [km/s]

Figure 3.6 Frequency - phase speed behaviour of the effect of gauge length averaging
in DAS recording: Gauge length: 20 m
 3.4 Gauge length issues 25

100.0 100.0
Gauge length 10 m Gauge length 20 m
90.0 90.0

80.0 80.0

70.0 70.0

60.0 60.0
Frequency [Hz]

Frequency [Hz]
50.0 50.0

40.0 40.0

30.0 30.0

20.0 20.0

10.0 10.0

0.0 0.0
0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00 0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00
Phase speed [km/s] Phase speed [km/s]

Figure 3.7 Representation of the differences between the first approximation (3.4.6)
and the full expression for gauge-length averaged strain rate (3.4.4). In the black area
the difference exceeds 10 percent of (3.4.6).

When the gauge length is expanded to 20 m, as shown in Figure 3.6, the

influence of the averaging of strain rate becomes more significant. Now
frequencies as low as 20 Hz are affected and the modulation starts to cut into
the observable range for near-surface studies.
Although there are arguments in favour of expanding gauge length to
optimise signal to noise, the use of longer gauge lengths brings with it larger
modulation effects that can have deleterious effects.

Low-frequency response
When the frequency is low the sine term in (3.4.4) can be expanded in a power
series so that
2ω ³ ωg ´ 2ω h ωg 1 ³ ωg ´3 i
〈ė d (ω)〉 = u sin ≈u − + ... . (3.4.5)
g 2c g 2c 6 2c
Hence in a first approximation the DAS response is a scaled version of the
ground acceleration.

ω2
〈ė d (ω)〉 ≈ u (3.4.6)
c
The difference between the full expression (3.4.4) and this first approximation
depends on ω4 and so grows rapidly with frequency

2ω ³ ωg ´ ω2 h 1 2ω ³ ωg ´3 i ω4 g 2
u sin −u ≈ −u = −u (3.4.7)
g 2c c 6 g 2c 24c 3
The difference (3.4.7) is plotted in Figure 3.7 for gauge lengths of both 10 and
20 m. For small phase speeds there is only a modest range of frequencies for
26 Representing the seismic wavefield for DAS

c: 500 m/s c: 1500 m/s c: 2500 m/s

fr: 75 Hz fr: 75 Hz fr: 75 Hz

0.05 0.10 0.05 0.10 0.05 0.10

Time [s] Time [s] Time [s]

fr: 50 Hz fr: 50 Hz fr: 50 Hz

0.05 0.10 0.05 0.10 0.05 0.10

Time [s] Time [s] Time [s]

fr: 25 Hz fr: 25 Hz fr: 25 Hz

0.05 0.10 0.05 0.10 0.05 0.10

Time [s] Time [s] Time [s]

Figure 3.8 Impact of gauge length: pulse shapes using the full DAS represntation (3.4.4)
shown in red are compared to the scaled acceleration form (3.4.6) shown in black as
a function of the reference frequency f r for gauge length 10 m and phase speeds of
500 m/s, 1500 m/s and 2500 m/s.

which the approximation (3.4.6) is effective. Regional and teleseismic phases

with steep paths at the surface can be well represented but the approximation is
much less suitable for surface waves and multiply reflected body wave phases.
Further, the sinsusoidal modulation of (3.4.4) at higher frequencies will alter the
shape of the waveform compared with scaled acceleration.
In terms of the wavelength of the seismic wave λ = c/ f the sinsusoidal term
takes the form
³ ωg ´ ³ πg ´
sin = sin , (3.4.8)
2c λ
so the behaviour is determined by the relative size of the gauge length and the
wavelength. The linear approximation works well for wavelengths longer than 6
times the gauge length, and waveform distortion only starts to appear when the
wavelength is shorter than 3 times the gauge length.
The impact of the full gauge length modulation as a function of frequency
is illustrated in Figure 3.8. Here we use the same style of pulse, a minimum
phase version of a Ricker wavelet, and look at both the full DAS response with
3.5 Gauge length effects for a bent cable 27

10 m gauge length in red compared with the slowness scaled acceleration form
appropriate for lower frequencies plotted in black. Pulse shapes are displayed
for a set of phase speeds (500, 1500 and 2500 m/s), and reference frequencies
(15, 50 and 75 Hz).
For a phase speed of 2500 m/s the full impact of gauge length is muted,
with little impact on pulse shape and just a loss of amplitude at the highest
frequencies. At the intermediate phase speed of 1500 m/s mild pulse distortion
has set in at 75 Hz reference frequency and the estimate of pulse onset might
be biased. However for the low phase speed of 500 m/s, the impact of the
gauge length is already significant at 25 Hz, since the wavelength is only twice
the gauge length. For the higher frequencies the pulse shape induced by
gauge-length effects is quite unlike the simple acceleration approximation, as
might be expected from Figure 3.7. Thus for even modest frequencies for waves
with low apparent phase speed, the the pulse shape with DAS recording is
radically altered and time picks of DAS records will differ from the true onset
time.

3.5 Gauge length effects for a bent cable

In many circumstances the fibre-optic cable being used for DAS registration
does not lie in a straight line. Particularly for dark fibre exploiting
existing telecommunications facilities the geometry is dictated by external
considerations.
Once again assume a plane wave from a distant source at distance ∆ with
slowness p The averaged strain rate over a gauge length is

1 x+g /2 iω h
Z ix+g /2
〈ė d (x, ω)〉 = ds ė(s, ω) = u(d.n)eiω(d.p)s eiω∆ , (3.5.1)
g x−g /2 g x−g /2

where u is the amplitude, d is a unit vector along the local cable direction and n
is the polarization vector for the plane wave.
Thus
iω h i
〈ė d (ω)〉 = u (d1 .n)eiω(d1 .p)g /2 − (d2 .n)e−iω(d2 .p)g /2 , (3.5.2)
g
where d1 = d(x + g /2), and d2 = d(x − g /2). We denote the inclination to the
cable, relative to the radial direction from the source, by ψ and so d1 .p = cos ψ1 ,
and d2 .p = cos ψ2 .
For waves oriented along the radial direction d1 .n = cos ψ1 , and for transverse
waves d1 .n = sin ψ1 .
Expanding the complex exponentials in (3.5.2) we have
³ ωpg ´ ³ ωpg ´
e+iω(d1 .p)g /2 = e+iω cos ψ1 g p/2 = cos cos ψ1 + i sin cos ψ1 , (3.5.3)
2
³ ωpg ´ 2
³ ωpg ´
e−iω(d2 .p)g /2 = e−iω cos ψ2 g p/2 = cos cos ψ2 − i sin cos ψ2 .
2 2
28 Representing the seismic wavefield for DAS

For radial polarisation

iω hn ³ ωpg ´ ³ ωpg ´o
〈ė d (x, ω)〉r = u r cos ψ1 cos cos ψ1 − cos ψ2 cos cos ψ2 (3.5.4)
g 2 2
n ³ ωpg ´ ³ ωpg ´oi
+i cos ψ1 sin cos ψ1 + cos ψ2 sin cos ψ2 .
2 2
For the low frequency case when ωpg /s ¿ 1, i.e., f ¿ 1/{πpg }, we can simplify
the frequency dependent terms
³ ωpg ´ ³ ωpg ´ ωpg
cos cos ψ1 ≈ 1, sin cos ψ1 ≈ cos ψ1 , (3.5.5)
2 2 2
and the expression for the averaged strain rate reduces to
iω © ª ω2 p ©
〈ė d (x, ω)〉r = u r cos ψ1 − cos ψ2 − u r cos2 ψ1 + cos2 ψ2 .
ª
(3.5.6)
g 2
Hence, in addition to a term proportional to the ground acceleration that
appears in the case of a straight cable, we also have a contribution proportional
to the ground velocity that depends on the differences in inclination at the ends
of the gauge length.
Similarly for the transverse component
iω hn ³ ωpg ´ ³ ωpg ´o
〈ė d (x, ω)〉t = u t sin ψ1 cos cos ψ1 − sin ψ2 cos cos ψ2 (3.5.7)
g 2 2
n ³ ωpg ´ ³ ωpg ´oi
+i sin ψ1 sin cos ψ1 + sin ψ2 sin cos ψ2 ;
2 2
with the low frequency approximation when f ¿ 1/{πpg },
iω © ª ω2 p ©
〈ė d (x, ω)〉t = u t sin ψ1 − sin ψ2 − u t sin ψ1 cos ψ1 + sin ψ2 cos ψ2 .(3.5.8)
ª
g 2
As expected, the expressions (3.5.6), (3.5.8) reduce to the familiar forms in
the case of a straight cable. These approximate expressions are expected to
work for teleseismic P waves up to 40 Hz, and for teleseismic S waves to around
20 Hz. The larger slowness for arrivals from regional sources means that the
suitable frequency range will be more restricted, but the approximations should
work well up to 10 Hz for P waves. For slight deviations for a straight cable the
additional effects will be small and unlikely to have any apparent influence on
observations.
In Figure 3.9 we illustrate the effect of a sharp bend in a cable, as for example
encountered when a telecommunications cable follows segments of orthogonal
streets. The recommendations for fibre optic cable laying require that the radius
of curvature at a bend be no less than 10 times the diameter of a cable. hence
for a 3 mm diameter cable the minimum radius of curvature would be 30 mm,
and commonly rather larger. If a cable loop is introduced a typical length would
be at least 1 m. However, a corner or a simple cable loop will be short compared
with the typical gauge length of 10 m employed for seismological applications.
3.5 Gauge length effects for a bent cable 29

(a) ψ1
cable
A C

Pla
ne
wa
ve
fro
nt
B
ψ2

(b)
Orientation factor

B
cos2 ψ2
C
~2g
A
cos2 ψ1

Distance along cable

Figure 3.9 Influence of a strong cable bend on DAS response: (a) At sample point A the
gauge length extends along a straight cable segment, and so also at B but now with a
very different inclination to the incoming plane wavefront; whereas for sample point C
the two ends of the gauge length sense different wavefront inclinations. (b) The orien-
tation factor for the mixed segment C is the average of the factors for the two straight
legs and persists for nearly two gauge lengths (2g ), in addition in this segment there will
be a ground velocity dependent term.

At each nominal sample point, the extracted strain-rate signal will represent
the average along the gauge length surrounding the point. Thus at A in Figure
3.9 the gauge length will span the zone between the red points with a wavefront
inclination of ψ1 . At B on the orthogonal leg, the inclination along the gauge
length will again be constant but now set at ψ2 . For a sample point such as C
the inclination of the incoming wavefront will be different at the two ends of the
gauge length and the full formula (3.5.6), (3.5.8) apply. The apparent inclination
factor will be (cos2 ψ1 + cos2 ψ2 )/2, but in addition there will be a contribution
proportional to ground velocity. The modified factors apply as soon as the
leading end of the gauge length enters the second segment and persist until the
trailing edge is on a the new straight trajectory.
The segment of the cable with rapidly varying inclination will be short
compared with the gauge length and so for a sequence of sample points at 2 m
intervals there will appear to be a sharp jump in the observed amplitude as the
30 Representing the seismic wavefield for DAS

sampling point enters the mixed regime, and again as it exits to the new state.
The mixed state persists for close to two gauge lengths and so the positions
of the geometry change can be estimated quite accurately as the midpoint
between the jumps in the character of the DAS records.
4
Modal fields

With the high density of sampling of the seismic wavefield achievable with DAS
recording it is frequently possible to extract higher modes of surface waves as
well as the fundamental mode. Seismic energy trapped between the free surface
and the steady increase in seismic wavespeeds with depth can propagate to
substantial distances. With dense sampling the full character of such guided
waves can be revealed. In favourable circumstances observations of partially
trapped waves such as ‘leaking modes’ and P-wave dominated features can be
tracked in addition to the more commonly studied S dominated fully-trapped
modes. As a result phenomena that previously have been regarded of minor
theoretical interest can be actually be observed. Understanding the nature of
such additional features aids improved rendering of structure at depth.
The character of the modal branches and their dispersion with frequency and
phase speed depends on interactions between different aspects of the seismic
structure and between wavetypes. The influence of P waves can become quite
strong in the very near surface when there are strong wavespeed gradients.
An effective tool for understanding the nature of the modal interactions
due to structure is provided by the seismic response in frequency–phase
speed space, the kernel for seismogram calculation. Such displays for all
three-components of the surface response extract the full modal response for
Rayleigh and Love waves, including leaking mode effects, and an indication of
the way that the modes are excited.

4.1 Surface waves

The advent of very high density observations of the seismic wavefield has
enabled the extraction of the dispersion of the fundamental mode and several
higher modes at relatively high frequencies (e.g., Chen et al., 2022, Fichtner et
al., 2023; Yang et al., 2024). Partially trapped P-dominated modes have also been
imaged in frequency–phase speed space (Li et al., 2022; Fichtner et al., 2023;
Yang et al., 2024).
The basic concepts of surface wave dispersion are well established, but
the improved observations require understanding of more complex modal

31
32 Modal fields

interactions at high frequency. Each of the major components of structure

near the surface have the capacity to trap seismic energy. As a result the
sedimentary zone, the crust and the uppermost mantle form separate but
interacting waveguides. The dispersion curves for modes with different
character apparently cross, whilst a single mode can switch its nature along its
trajectory in frequency – phase speed space (e.g., Kennett, 2023a).
Although the properties of Rayleigh waves are dominantly controlled by the
behaviour of SV wavespeed with depth, where low P wavespeeds occur near
surface there can be significant interaction between SV - and P-dominated
modes. The nature of these waveguide interactions can be understood in terms
of the controls on modal dispersion.
In a full three-dimensional body, such as the Earth, all deformations can
be completely described by a superposition of the set of normal modes of the
structure. However, when we consider a truncated portion of the structure
in cartesian geometry, the properties at depth set a limit on the range of
phase speeds for which surface wave modes are fully trapped. Waves with
higher phase speed than this limit lose energy as they propagate horizontally by
radiation into the lower halfspace, but can sometimes be recognised as ‘leaking’
modes extending from the trajectory of fully trapped modes.
For a fully elastic medium, the trapped modes of the regular surface waves
correspond to poles of the response lying in the real frequency-slowness plane.
When radiation leakage occurs, the singularities for the leaking modes occur on
lower sheets in complex slowness at each frequency (e.g., Haddon 1984). In the
presence of attenuation all poles are shifted into the complex plane. However,
for typical levels of seismic attenuation the dispersion trajectories of the poles in
frequency-slowness space can be retrieved with a purely elastic approximation.
For a realistic seismic structure, including attenuation, the character of the
modal field can be extracted by direct calculation in the frequency-slowness
domain with modal branches defined by localised increases in amplitude
reflecting the presence of poles off the top slowness sheet (Kennett 2023a).
In this way both trapped modes and leaking modes can be visualised, and
the effects associated with the interactions of multiple seismic waveguides
analysed. This approach avoids the numerically intensive exercise required to
track leaking mode singularities on to lower sheets in complex slowness space
With individual seismometers deployments were commonly made on hard
rock to improve signal to noise, but dense deployments are frequently placed
in areas with sedimentary cover with lower wavespeeds than the basement.
From the basement through the crust seismic wavespeeds increase steadily,
and then increase again on entry to the upper mantle at the Moho. Through
the uppermost mantle the wavespeed increases again though more slowly,
with distinct jumps at the upper-mantle discontinuities near 410 and 600
km depth. Below the sediments wavespeed gradients are generally not
4.2 The nature of the modal field 33

strong, but are sufficient to turn back energy to the surface. When working
in cartesian geometry, compensation for the effects of sphericity is made
with earth-flattening transformations that increase the vertical gradients to
compensate for the decrease in horizontal distance scale with depth (e.g.,
Chapman, 1973).
The sediment, crustal and mantle zones each have a linked set of S
wave dominated reverberations from the free surface whose interference
controls associated surface wave dispersion. However, multiple reflections
with the same slowness (phase speed) can span multiple zones and so tie
together the different dispersion characteristics. This means that an individual
surface wave mode in different slowness ranges will have a character tied to
different waveguides with link points where mode branches nearly meet. The
presence of low velocity channels within the crust and mantle also produces
further opportunities for guided waves and transitions between channelled
contributions and modes with the character of the main structure (e.g., Kerry,
1981).

4.2 The nature of the modal field

Consider an stratified elastic medium with surface P and S wavespeeds α0 , β0 ,
underlain by a uniform halfspace with properties αL , βL . The response of the
medium to a surface source in the frequency–slowness (ω − p) domain from
(3.2.1) takes the form

w0 = WF [I − R0L −1 0L
D RF ] RD (4.2.1)

where R0LD is the reflection response of the entire structure below the surface,
RF the matrix of free-surface reflection coefficients, WF the surface response
including free-surface effects and I the identity matrix. Here the source has
been left arbitrary. Different aspects of the seismic wave field can be isolated
by selecting the components of R0L D to be modulated by the reverberations
in the structure (the inverse term) and the free surface response WF . The
contributions RF and WF depend only on slowness (phase speed), whereas R0L D
is a function of both slowness and frequency.
For isotropic, and transversely isotropic, media the response (4.2.1)
breaks into two distinct parts. SH waves propagate independently and
the characteristics of the medium are expressed through the SH reflection
coefficient [R0L
D ] H H . The P and SV components are coupled, so that 2×2
matrices of reflection coefficients and free-surface terms are needed
µ PP PS¶ µ PP PS¶ µ ZP
WF WFZ S
¶
0L RD RD RF RF
RD = SP SS , RF = SP SS , WF = , (4.2.2)
RD RD RF RF WRPF WRSF

with a compressed notation so that

RPF P = [RF ]P P , RPDS = [R0L

D ]P S , WFZ P = [WF ] Z P etc. (4.2.3)
34 Modal fields

The subscript Z indicates the vertical component and R the radial component.
The P wave response of the structure can be emphasised by selecting just the P P
and P S elements of the matrix R0L D in (4.2.1) and the SV response by selecting
the SS and SP elements.
The response of the medium specified by (4.2.1) will be singular when the
determinant of the reverberation term vanishes, i.e., det[I − R0L
D RF ] = 0. The
locus of such values in frequency-slowness space determines the trajectory
of the dispersion branches for seismic modes. Here we concentrate on the
character of such trajectories in the frequency–phase speed domain.
Love waves – SH
The dispersion relation for the entire structure takes a relatively simple form for
Love waves, involving only the SH (tangential) component:

1 − [R0L
D (p, ω)] H H = 0,
¡ ¢
(4.2.4)

since SH waves do not couple to P and the free-surface reflection coefficient for
SH waves is unity. The dispersion condition (4.2.4) requires the SH reflection
coefficient for the entire structure to be equal to unity and so its phase χD (p, ω)
must be a multiple of 2π:

χD (p, ω) = 2nπ, n = 0, 1, 2, . . . (4.2.5)

This property can be used to set up an efficient recursive system for models
without localised low velocity zones (Kennett & Clarke, 1983).
Rayleigh waves – P-SV
For the P-SV case the dispersion relation for Rayleigh waves takes the form

det I − R0L
D (p, ω)RF (p) = 0,
¡ ¢
(4.2.6)

where RF (p) is the free-surface reflection matrix that depends on slowness p

alone and R0L D (p, ω) is the reflection matrix from the full stratification. Other
equivalent representations exist where the structure is broken at different levels
(Kerry, 1981; Chen, 1993).
In terms of the components of the free surface reflection matrix RF and the
reflection matrix R0LD from the structure beneath the surface, the dispersion
relation (4.2.6) can be written as

[1 − RSS SS SP P S PP PP P S SP
D RF − RD RF ][1 − RD RF − RD RF ]
−[RSP PP SS SP PP PS P S SS
D RF + RD RF ][RD RF + RD RF ] = 0. (4.2.7)

As noted in Chapter 16 of Kennett (2001) the form (4.2.7) does not represent a
physically feasible set of wave propagation processes, because we have P legs
juxtaposed with S wave legs, but (4.2.7) can be rearranged to represent a suite of
actual processes. We introduce the P-wave reverberation term

[X P P ]−1 = [1 − RPDP RPF P − RPDS RSP −1

F ] , (4.2.8)
4.2 The nature of the modal field 35

which represents all processes which involve multiple P waves, including the
possibility of wavetype conversion both at the surface and from the structure.
Dividing (4.2.7) by X P P and rearranging the terms

1 − RSS
© SS
+ RSP [X P P ]−1 RPDP RPF S + RSP [X P P ]−1 RPDS RSS
ª
D R F F F F
−RSP
© PS PP P P −1 P P P S
] RD RF + RPF P [X P P ]−1 RPDS RSS
ª
D RF + RF [X F = 0. (4.2.9)

Now each of the expressions in braces represent a physically realisable sequence

of propagation processes linked to the reflections from the structure contained
in RSS
D
, RSP
D
. The form (4.2.9) for Rayleigh wave dispersion indicates the
significant role that P-wave propagation can play in determining the dispersion
characteristics.
Kennett (2023a) has provided a detailed analysis of the nature of the Rayleigh
wave dispersion relation (4.2.6) across the different bands in phase speed.
Trapped modes arise for phase speeds lower than the S wavespeed in the
underlying halfspace βL down to the surface S wavespeed β0 , with only the
fundamental Rayleigh wave able to have lower phase speed – but still less than
the Rayleigh speed for a uniform half space with the surface properties. P waves
are evanescent for phase speeds less than α0 and then have only a modest
influence on Rayleigh wave dispersion.
When the surface P wavespeed is less than βL it is possible for trapped
modes to involve both P and S reverberations with two interacting waveguides.
Even in this regime the main Rayleigh mode branches are controlled by the
influence of S-wave structure, with a spacing of branches at a given slowness
approximately proportional to the inverse of the delay time for reflection from
depth. For P-dominated modes this delay is roughly half the size and so their
spacing is much larger than for the S-dominated modes and their slope in
the frequency–phase speed space is steeper. As a result their trajectories cut
across the regular Rayleigh modes. The character of a single mode branch will
switch with phase speed according to which dispersion relation is most closely
satisfied. There is a rapid change in slope near where both P and S dispersion
terms approach zero and in such zones successive mode branches nearly touch
(‘osculation’ points).
For phase speeds higher than βL , S waves radiate into the lower half-space,
and the S-dominated branches head into the leaking mode regime on lower
sheets in the complex slowness plane at each frequency. For phase speeds just
above βL there is a muted influence on the top sheet from the proximity of the
poles. With phase speed still lower than the P wavespeed in the underlying
halfspace there is still partial trapping for P waves with loss occurring on
conversion to SV waves at each surface reflection. With a strong P wavespeed
gradient such partially trapped waves can persist to considerable distance and
the associated complex poles of the response (4.2.1) can have a noticeable effect
on the top sheet.
36 Modal fields

Frequency

Phase velocity

Figure 4.1 Schematic representation of multiple Rayleigh modes associated with the
SV waveguide (green), apparently crossing the modal curves for the P waveguide with
higher wavespeeds (red). The spacing of modes is inversely proportional to delay time
and so larger for the P-dominated modes. The modes branches do not actually cross
but approach closely at the osculation points.

4.3 Interacting waveguides

The full dispersion relation for Rayleigh waves can be written in the form

[1 − RSS SS PP PP
D RF ][1 − RD RF ] −C
SP
= 0. (4.3.1)

where the term C SP involves one or more conversions between wavetypes. If

dispersion was controlled solely by S wave structure then [1 − RSS D F
RSS ] would
PP PP
vanish and similarly if just P propagation was involved [1 − RD RF ] must
vanish. When both P and SV waves are propagating within the stratified
structure there is a competition between the influences of the two dispersion
terms for different classes of waveguide.
For fixed phase speed c = 1/p, consider the neighbourhood of a root of
[1 − RSS
D F
RSS ] = 0 at frequency ωS . The actual root for (4.3.1) will lie close at
ω = ωS + ∆ω and so
∂ ¯
[1 − RSS
D FRSS
] ≈ ∆ω [1 − RSS SS ¯
R ]
D F ¯ω (4.3.2)
∂ω S

From (4.3.1) we then find

" #
C SP
∆ω ≈ ¡ . (4.3.3)
[1 − RPDP RPF P ]∂[1 − RSS RSS ]/∂ω + ∂C SP /∂ω
¢
D F ωS

Unless 1 − RPDP RPF P ] itself approaches zero, the frequency shift ∆ω will be small.
Hence the dispersion characteristics will be close to that for the SV case alone,
i.e., regular Rayleigh modes.
A comparable effect arises near frequencies that satisfy the P waveguide
relation [1 − RPDP RPF P ] = 0, but now involving the frequency derivative for the
4.4 Multi-mode surface waves and leaking modes 37

SV term. For such a coupling scenario we therefore have two sets of largely
independent mode segments satisfying the separate dispersion relations, with
strong interaction only where both dispersion relation are close to being
satisfied.
The regular Rayleigh modes will be linked to SV waves with longer delay times
and so relatively close spacing with low slope. The P dominated modes have a
shorter delay time and thus more widely spaced mode branches in frequency
with steeper slope. The separate suites of mode branches do not actually touch
but approach each other very closely at osculation points where the dispersion
conditions for each separate waveguide are very nearly satisfied. If we track
a single mode with phase speed we find that it will change its character as it
passes these points with alternating sections associated with the different mode
classes.
As discussed in Kennett (2023a) a comparable situation arises when the
structural model is split into a shallower and a deeper part. There will be a
mode branch suite with relatively wide spacing associated with the shallow
component where the delay time is small, these will appear to cross the modes
linked to the deeper structure with longer delay times and closer spacing. With
the addition of additional structural components with depth,there will be suites
of apparently distinct modes associated with each part of the structure that
interact strongly locally.

4.4 Multi-mode surface waves and leaking modes

For Rayleigh waves, the decomposition of the dispersion relation for each part of
the structure will have a strong variation with slowness because of the presence
of both S- and P- dominated modes. Thus, superimposed on the main mode
structure controlled primarily by the S wavespeed distribution, we have further
effects linked to the (partial) trapping of P waves.
For Love waves in an isotropic model, the situation is simpler there is only a
single wavetype and there is no slowness variation of the free-surface reflection
coefficient. Nevertheless, there is still the potential for interaction between
waves dominantly propagating in different parts of the stratification.
Each discontinuity in wavespeed structure or zone of strong wavespeed
gradients will have its own modest influence on the modal structure. In
frequency-phase speed space the presence of a discontinuity induces an
inflection in the slope of the modal trajectory around the wavespeed associated
with the faster layer, even when modal branches do not approach closely. In
effect, at each discontinuity we are seeing the interaction of the waveguides
above and below the interface. The largest effects are seen for major jumps in
physical properties, as in models comprised of a few uniform layers. Finger &
Löer (2024) have made use of the changes in surface ellipticity of ground motion
for multimode Rayleigh waves at such inflection zones to map out the depth of
major discontinuities
38 Modal fields

Strong effects on modal dispersion arise in the presence of pronounced

zones of lowered seismic wavespeeds within the structure that act as strong
waveguides, particularly at high frequencies. Channelled waves can be trapped
in such waveguides, with only weak linkage to the external structure via
evanescent waves (Kerry, 1981). The transition between the regular surface
wave modes and the channel waves is very abrupt as a function of slowness (or
phase speed) and can cause substantial difficulties in tracking individual mode
branches.
In principle, the dispersion characteristics of the various mode branches
can be determined by finding the roots of the equation (4.2.1) as a function
of frequency and phase speed. In practice, numerical root finding for
perfectly elastic models can become very tricky, with large numbers of different
mode branches at even moderate frequencies and intricate behaviour around
osculation points. Leaking modes present further difficulties since the singular
points no longer lie on the top sheet of the complex phase speed plane.
Such complications can be avoided by looking directly at the structural
response in the frequency–slowness or frequency–phase speed domain. Such
an approach captures all the different mode branches and their relative
excitation (e.g., Kennett, 2001, Chapter 16; Dal Moro, 2020). It is also possible
to look at polarisation issues. A major advantage of employing the full
frequency–phase speed spectrum is that all modes that make an effective
contribution to the wavefield can been seen, with the inclusion of realistic
attenuation. The trajectories for trapped modes appear as narrow zones of
elevated intensity in the response. Leaking modes still have an influence on the
top sheet with real phase speed values and so can be tracked by their associated
broader intensity zones.
Such spectral images have been employed by a number of authors,
particularly in the context of analysing models with a few uniform layers.
However the merits of the spectra are even more pronounced for more complex
models. A common formulation can be used to look at the separate behaviour
of the SH and P-SV systems, and the way in which the different components of
the wavefield separate modes with dominant SV or P character. Such diagrams
have proved very helpful in studies of near surface ice properties from DAS
records in Greenland (Fichtner et al., 2023) and at the South pole (Yang et al.,
2024).
From (4.2.1) the three components of the seismic response for a surface
source can be evaluated on a preassigned grid with even sampling in frequency
and slowness or phase speed. The use of phase speed provides greater
separation of the different aspects of the modal field and a more direct relation
to the wavespeed structure and so is easier to interpret.
An effective rendering of the nature of the wavefield can be achieved by
combing the three components in a single diagram. The tangential component
4.4 Multi-mode surface waves and leaking modes 39

β,α [km/s] ρ [Mg/m3]

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0
0.0

0.1

0.2
Depth [km]

0.3

β ρ α
0.4

0.5

0.6

Figure 4.2 Structural profiles for a shallow sedimentary basin. The P wavespeed (α) is
shown in red and the S wavespeed (β) in green, with the associated density (ρ) profile
in pale blue.
11.0

10.0

9.0

8.0

7.0
Frequency [Hz]

6.0

5.0

4.0

3.0

2.0

1.0

0.0
0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00
Phase speed [km/s]

Figure 4.3 Frequency – phase speed spectrum for the wavespeed model shown in Fig-
ure 4.2. The Love wave modes form the tangential component (green). For the Rayleigh
waves, the radial component dominated by SV waves is shown in blue and the verti-
cal component dominated by P waves in red. The osculation points associated with a
switch from S to P domination appear clearly for phase speed around 2.25 km/s.
40 Modal fields

with Love waves is placed on the green channel. For Rayleigh waves, the radial
component is placed on the blue channel and the vertical component on red. By
this means the portions of the frequency – phase speed spectrum with strongest
P wave influence appear in red.
For situations such a sedimentary deposits with strong gradients in
wavespeed the mode spectrum is rich and higher modes can be excited by near
surface sources. In Figure 4.2 we show a sedimentary structure for a shallow
sedimentary basin, with the associated frequency – phase speed spectrum in
Figure 4.3. The S wavespeed in the underlying half-space is 2.511 km/s and this
is reflected in the relatively sharp cutoff of the higher modes, especially for Love
waves. For Rayleigh waves, the first higher mode shows a strong extension as a
leaking mode, where the complex dispersion pole lies close to the top sheet of
the phase-speed plane. The leaking mode components become progressively
weaker with increasing mode number as the dispersion poles on the lower
complex sheets separate more rapidly from the top sheet.
Even though P waves are evanescent for small phase speeds, they have some
influence on the dispersion for the fundamental and first higher Rayleigh mode
that have a significant presence on the vertical component. The P wavespeed
profile overlaps the zone in which S waves are trapped and so parts of the higher
Rayleigh mode branches have P wave character. These portions appear in red
with steeper slope than the branches dominated by S waves (blue). As noted
above the mode branches for this isotropic model do not cross but approach
closely at osculation points where there is a rapid change of character. This
effect is picked up very clearly by the colour coding of the components.
The P-dominated branches extend beyond the S wavespeed cutoff as leaking
modes whose character is comparable to acoustic waves. P wave reflection at
the free-surface is not complete with loss to converted SV waves, and so the
dispersion poles lie at complex phase speeds. As noted above, the separation
of the mode branches is largely controlled by the inverse of the delay times
associated with reflection of the associated wavetype. The shorter delay times
for P reflection means that the mode branch separation is much larger. The
slopes of the P mode branches are comparable with those of low order Rayleigh
modes at the same frequency but shifted to higher phase speeds, where they are
steeper than those of the SV -dominated modes.

4.5 Modal spectra from DAS recording

As an illustration of the detailed rendering of the seismic wavefield made
possible with the dense sampling provided by DAS, we show results from
the recording of a landing aircraft at the EastGRIP drilling site in Greenland
(Fichtner et al., 2023). The landing of a Hercules aircraft was recorded along a
3 km cable approximately perpendicular to the landing direction (Figure 4.4).
The complex signals in the time and space domain were transformed into
frequency–wavenumber space by Fourier transformation, with allowance for
4.5 Modal spectra from DAS recording 41

Figure 4.4 DAS recordings from a landing aircraft at the EastGriP drill site in Greenland
(10 m gauge length). The configuration of the DAS cable and the Landing strip is shown
in the upper panel. Below, the left panel displays the time-domain DAS recording along
the straight section of the cable between 500 and 3000 m. The right panel shows the
spectrum in a frequency–phase speed representation. The most prominent features
include several Rayleigh wave modes, a leaky mode, and pseudoacoustic P-modes.
[After Fichtner et al., 2023].

the mild angular correction. Wavenumber was then further mapped into phase
speed to give a rendering of the frequency – phase speed spectrum (Figure 4.4).
The near-surface structure is built up from the accumulations of snow each
year and shows a strong gradient in wavespeeds as these contributions compact
into solid ice. This leads to strong excitation of higher Rayleigh modes by a
surface source, and also to trapping of P wave energy. The net result is a rich
spectrum of modes with clear indications of modal cutoff for both the Rayleigh
modes and the P modes at the wavespeeds corresponding to deep ice once the
compaction is complete.
Modelling of the spectra suggests that it is most likely that the feature labelled
R1 in Figure 4.4 is in fact the second higher Rayleigh mode. The first Rayleigh
overtone is only weakly excited and lies between the strong fundamental mode
R0 and the second overtone. With conventional seismic recording there is rarely
any indication of leaking modes continuing beyond the shear cutoff, but here
we see a clear continuation of the R1 feature as a leaking mode – a result of the
unaliased rendering of the wavefield.
42 Modal fields

As we saw in §3.4 there is strong influence of gauge length on the frequency

– phase speed behaviour for high frequencies and small phase speeds. For this
ice case the natural attenuation in the near-surface is sufficient to suppress the
modal branches before they enter the zone of complications (cf. Figure 3.5).
5
Near field

A important application of DAS systems has come from monitoring close to

sources. This has been done with both specially laid cables and active sources
as well as the use of ‘dark fibres’ in existing telecommunications cables.

For all experiments that make use of existing cables there is a need to
carefully map out the geometry so that recording channels can be directly
assigned. Although the general pattern of cable run may be known, the distance
along the ground is commonly shorter than the distance along the fibre being
interrogated, due to slack in the fibre deployment. Additional effects can also
arise where cable loops are included at access points.

The most common approach is to use localized sources (tap tests) to calibrate
channel count against position. The source may be as simple as a dropped
weight or hammer blow at a calibrated point, with typically the assigned
channel being that with the least time offset. Even in this case, it is hard to
reduce the calibration error to half the gauge length. In an urban environment,
such sources may present difficulties and an alternative has been sought in
multiple jumps on a sidewalk beside the road beneath which the cable passes
(e.g., Smolinski et al., 2024). For complex geometries with many changes in
cable direction, an indication of the position of cable bends can be extracted
from the modulation of the DAS strain-rate amplitude for distant events due to
orientation effects (§3.5).

An alternative for long lengths of dark fibre is to map out channel locations
on cables under or alongside roads using the DAS response to a moving vehicle
(Biondi et al., 2023) with a common time base for the vehicle navigation and the
DAS recording system. The weight of a car moving at a speed slower than the
surface Rayleigh waves produces a localized strain-rate signal for frequencies
below 1 Hz (e.g., Yuan et al., 2020). With low pass filtering this quasi-static signal
can be separated from the seismic waves generated by the passage of the vehicle
and employed to identify the closest DAS channels to within a gauge length.

43
44 Near field

Figure 5.1 Example of DAS recording using a dark fibre along streets in Bern,
Switzerland. The optical intensity is proportional to strain rate. Marked features: (a)
Movement of a vehicle along the fibre, appearing to induce surface waves; (b) A second
vehicle signal; (c) response of leaf blower. The fibre runs alongside a water main in the
distance range 300–600 m, producing a consistent background signal in this area. The
DAS strain-rate amplitudes are consistently reduced in the distance ranges 20–100 m
and 230–280 m, that correspond to fibre access locations. Occasional bursts of signal
are seen (e.g., around 6 s) within these gaps, which could be caused by a vehicle driving
over the manhole. The channels from 750–1200 m are consistently noisy because the
fibre runs beneath many large buildings with strong local activity. [Courtesy of Krystyna
Smolinski].

5.1 Urban noise environment

As in all branches of seismology effective use of DAS records requires separation
of the desired signal from extraneous noise. Particularly where existing cables
5.1 Urban noise environment 45

Figure 5.2 A 30 minute sample of urban DAS data from a 450 m section of fibre,
showing the movement of vehicles along the main Bern street. The amplitude
corresponds to strain rate. [Courtesy of Krystyna Smolinski].

Figure 5.3 Example of ‘fish-bone’ propagation pattern produced by a car. The

‘backbone’ shows the velocity of car moving along the fibre; the more steeply-inclined
‘bones’ are the induced Rayleigh waves, propagating with much higher phase speed.
[Courtesy of Krystyna Smolinski].

are being utilised, the surrounding noise environment can be troublesome, with
strain-rate disturbances due to vehicles or even trains impinging on the fibre at
the same time as desired signal.
As an illustration of the urban noise scenario, Figures 5.1–5.3 show examples
of short time sequences of DAS records from a fibre connecting educational
institutions in Bern, Switzerland, mostly lying beneath a well-travelled road
(Smolinski et al., 2024). In Figure 5.1 a 10 second time window is shown for the
entire cable with a number of different classes of features. The marked sources
arise from vehicles and from a leaf blower employed near the side of the road.
Consistent gaps in the DAS response are associated with cable access locations,
and may involve looped cable. Minor modulations of the DAS amplitude in
the distance interval 1200–2100 m are likely to be linked to cable slack at a
number of access points along the road. Figure 5.2 shows a longer term view
of the passage of vehicles past the DAS cable in Bern. The dominant signal
from a vehicle is in the form of Rayleigh waves shed from the moving source.
Stronger, slower contributions are associated with buses and delivery vehicles.
46 Near field

(a) Strain (b) Strain

1.0 1.0
0.8 0.8
0.6 0.6
0.4 0.4
0.2 0.2

-12 -8 -4 0 4 8 12 -12 -8 -4 0 4 8 12
Distance [m] -0.2 Distance [m] -0.2
-0.4 -0.4
-0.6 -0.6
-0.8 -0.8
-1.0 -1.0

(c) Strain (d) Strain

1.0 1.0
0.8 0.8
0.6 0.6
0.4 0.4
0.2 0.2

-12 -8 -4 0 4 8 12 -12 -8 -4 0 4 8 12
Distance [m] -0.2 Distance [m] -0.2
-0.4 -0.4
-0.6 -0.6
-0.8 -0.8
-1.0 -1.0

Figure 5.4 Influence of gauge length and instrumental averaging on simple signals
with a narrow span: (a),(c) half-width 1 m and (b),(d) half-width 3 m. The input strain
components are shown with solid lines, the effect of 10 m gauge-length integration by
chain dotted lines and the DAS output at channels separated by 2 m as solid lines with
markers at the channels. This last stage includes instrumental stacking over 2.75 m
around the output channel, combining 11 signals at 0.25 m separation. Red traces are
for a symmetric strain signal about the origin, and blue traces for an antisymmetric
strain signal.

Detail of the DAS signal from a car is shown in Figure 5.3), the steady signal
comes directly from the moving vehicle which also sheds Rayleigh wave trains
as it moves.

5.2 Taptest effects

The effective calibration of DAS channel locations is based on the response
to sources close to the location of the cable. In the immediate vicinity of a
such a source, there are inevitably rapid variations in the nature of the seismic
wavefield that are tamed and somewhat modified by a DAS recording system.
Most DAS signal comes from along-cable strain but can be augmented slightly
by cross-cable effects.
For sources close to the DAS fibre, the records are particularly sensitive to the
5.2 Taptest effects 47

internal processing in the interrogator. In addition to the spatial averaging due

to the gauge length, channel stacking is employed to counteract optical fading
issues. The influence of such effects are illustrated in Figure 5.4 for symmetric
and anti-symmetric strain signals with half-widths of 1 m and 3 m for a gauge
length of 10 m and output channels at 2 m spacing. Simple stacking over 11
subchannels at 0.25 m spacing is employed centred on the each of the output
channels.
For a symmetric strain signal the combination of gauge-length averaging and
stacking acts to spread the apparent effect (Figure 5.4), so that even for a narrow
input the DAS signature will extend over at least the full gauge length. The shape
of the resultant DAS response is only weakly dependent on the nature of the
input strain signal.
For a ground velocity field that is symmetric about the origin, the
corresponding strain will be antisymmetric. This antisymmetry about the origin
is preserved in the gauge averaging, but the main DAS features are displaced
from the original strain concentration, especially when this is narrow.
Although the seismic wavefield is at its most complicated in the immediate
vicinity of a source, Kennett et al. (2024) have shown that is possible to get
a good indication of the expected behaviour for tap tests using a simple ray
based approach, with a locally uniform medium, in combination with the
radiation characteristics for different sources. The near-surface zone tends to
be somewhat heterogeneous in physical properties, but for short propagation
distances the distortion of wavefronts by local structure will be limited and
so useful results can be obtained with the use of a uniform medium without
needing to know the actual conditions in detail.
For a DAS cable beneath the source point the strain signal is symmetric for a
vertical force and anti-symmetric for a horizontal force, so that from Figure 5.4
the DAS response is more complex when horizontal forces are present. With a
hammer or weight drop source, horizontal forcing can arise when the impact
on the ground is not purely vertical. Even with a vertical impact, if the ground
surface is sloping and the cable run follows the slope, an apparent forcing
component parallel to the cable will have the same character as for a horizontal
force at a flat surface (Kennett et al., 2024). The action of the gauge-length
averaging and subsequent stacking in the DAS system mean that the net effect
near the source depends only weakly on the distance from the source to the
cable.
In urban situations, a common scenario is that the cable with the fibre being
used for DAS measurements is placed at relatively shallow depths under, or to
the side of, a road. Tap tests are then likely to have a horizontal offset from the
cable, and so there will be Rayleigh waves generated by the source in addition
to P and S body waves. Unless the distance from the source to the cable is
comparable to the gauge length the main peak in DAS response will occupy
48 Near field

Site
ch434

ch433

ch432

ch431

ch430

ch429

ch428

ch427

ch426

ch425

ch424

ch423

ch422

ch421

ch420

ch419

ch418

ch417

ch416

ch415
27
Time [s]

27.9

28.0
Site
ch553

ch552

ch551

ch550

ch549

ch548

ch547

ch546

ch545

ch544

ch543

ch542

ch541

ch540

ch539

ch538

ch537

ch536

ch535

ch534
31
35.1
Time [s]

35.2

-20.0 -10.0 0.0 10.0 20.0

Distance [m]

Figure 5.5 DAS records for tap tests at sites 27 and 31 at the Tidbinbilla site, with a
weight drop source about 2 m above the DAS cable. Gauge length 10 m.

approximately the gauge length, flanked by weaker zones with reversed sign.
In scenarios where roads are below-ground structures, such as cellars, surface
waves can be reflected from close to the source and arrive as an extension of the
coda of the source or a distinct later arrival.
The expectation for simple tap tests is that the strongest DAS signal, and
earliest time arrivals, will occur at the recording channel closest to the source.
As we have seen the effect of DAS processing means that the main signal will
be sustained over at least a gauge length, but the midpoint of this feature is
normally the best estimate.
Nevertheless, the actual situation can differ from the usual scenario where
sites show local topography. Figure 5.5 shows results from two weight-drop tap
tests above a cable at about 2 m depth beneath the surface. A 1.9 km dark fibre
at the Tidbinbilla Deep Space Tracking Station south of Canberra in Australia
was used, which connected from the main building to the instrument room of
an temporarily inactive dish. Multiple weight drops were made at a number of
calibration locations with good reproducibility of DAS records. Site 27 lies on a
sharp slope with about a 15% gradient approximately in line with the cable run,
whereas Site 31 has a weaker gradient (about 7%) inclined at a angle or roughly
5.2 Taptest effects 49

D0228

D0227

D0226

D0225

D0224

D0223

D0222

D0221

D0220

D0219

D0218

D0217

D0216

D0215

D0214

D0213

D0212

D0211

D0210

D0209
1.0
Site
A
Time [s]

1.5

Site
D1315

D1314

D1313

D1312

D1311

D1310

D1309

D1308

D1307

D1306

D1305

D1304

D1303

D1302

D1301

D1300

D1299

D1298

D1297

D1296
B
1.0
Time [s]

1.5

-20.0 -10.0 0.0 10.0 20.0

Distance [m]

Figure 5.6 DAS records for jumps at Site A and B in Bern, lying about 2 m away from
the DAS cable. The shaded area highlights the spatial extent of the late coda most likely
due to reflected Rayleigh waves from below-ground structures at the sides of the street.
Gauge length 10 m.

30◦ to the cable. At Site 27 (FIgure 5.5), the DAS records are anti-symmetric
about channel 425 indicative of a significant force component parallel to the
cable. This arises from the effect of the ground slope on the vertical weight drop
and the slope of the cable that is presumed to follow the surface trend. For Site
31 the impact of the slope is smaller but still the DAS response is asymmetric,
and the putative calibration channel could be displaced from the closest by a
few metres.
The complexity of the built environment in urban settings can produce
quite complex influences on the records from tap tests. Figure 5.6 shows two
examples of records from an experiment along a dark fibre in Bern, Switzerland
lined with buildings with cellars. The source was an individual jumping, and
successive jumps gave comparable records. The signal from the jumps carried
to considerable distances, about 40 m from the source, but in a number of
instances produced rather complex DAS records over a zone a little larger than
a gauge length around the apparently closest channel. The extended coda and
distinct later arrivals appear to be a product of the 3-D structural situation,
50 Near field

with reflected Rayleigh waves from sub-surface concrete walls. The DAS gauge
length of 10 m is quite long compared to the distance from the jump point to the
nearest building cellars and so even within the gauge length reflections occur
at relatively wide angles. Strong reflection is to be expected at near-normal
incidence but its effect will be spread by the DAS strain-rate averaging over at
least the gauge length.
The action of gauge-length and stacking effects is to broaden the zone of
highest amplitude DAS signal in the neighbourhood of the source location, as
we have seen for simple human-powered sources. Comparable issues arise
when vehicle sources are employed, and the efficacy of calibration will be
improved when a car travels on the segment of the road closest to the cable line.
For dark fibre such tuning will depend on how closely the original fibre layout
has been mapped.
An additional issue with sources very close to a cable is that strain rates
may be such that the relative optical phases of the returned pulses cannot be
correctly measured. In this case the apparent DAS signal is distorted (‘clipping’).
The dynamic range of DAS systems is large, but deleterious effects can occur
when the differential phase is close to π or −π, because of the 2π ambiguity in
phase and consequent difficulty in phase unwrapping (Katakami et al., 2024).

5.3 Local wavefields

Where local structure is approximately stratified, simulations of DAS records
using the frequency–slowness representations developed in Chapter 3 can be
effective in understanding both source and structural issues.
To be able to model DAS records we must find
(i) a suitable source representation,
(ii) an appropriate propagation model and
(iii) a detailed description of the configuration of the channels along the
optical fibre in terms of position and orientation.
For the local case we need to include both near- and far-field components of
the seismic response. For a horizontal cable, the averaged axial strain rate along
a gauge length is best treated in terms of the differences in ground velocity
resolved along the cable at the ends of the gauge length (3.1.1). This differencing
can be readily implemented directly in the frequency–slowness domain (cf.
3.1.3) using a mild modification of existing modelling procedures for stratified
media (e.g., Kennett, 1983, 2001), assuming cylindrical symmetry about the
source point.
It is simplest to produce single channel DAS responses in the simulations,
with post-processing if needed to provide a direct match to specific DAS
interrogator settings. Any simple simulation has to assume a constant coupling
factor for a DAS cable to its environment. Amplitude discrepancies from
modelling can help reveal variations in cable coupling. However, deviations
 5.3 Local wavefields 51

(a) DAS strainrate (b) Modelling

40 40
D D

3
30 30

(c)
β,α [m/s]
20 20
0 200 400 600 800 1000

10
2
10 10
Distance [m]

Distance [m]
20
2
1 30
1
0 0

Depth [m]
40

50

-10 -10 60

70

-20 -20 80

90
β α
-30 -30 100

-40 -40

0.00 0.10 0.20 0.30 0.40 0.50 0.00 0.10 0.20 0.30 0.40 0.50
Time [s] Time [s]

Figure 5.7 A jump test in a street in Bern, Switzerland: gauge length 10 m, channel
spacing 2 m. (a) Strain rate observations from Silixia iDAS with 1000 Hz sampling. (b)
Simulation by velocity differencing in the slowness-frequency domain. The synthesis
used 2500 slownesses out to 4 s/km−1 , for an 0.8192 s time interval with 4096 time
points. The pass band was 0.25 to 55 Hz and moderate attenuation was included in the
model with Q p−1 = 0.01, Q s−1 = 0.02. Travel time calculations for a source displaced 2 m
from the cable are superimposed: P wave arrivals are shown in red and S wave arrivals
in green; D indicates the direct and refracted arrivals, the numbers indicate reflections
from the interfaces. (c) Wavespeed depth profile for P (red) and S (green) used in the
simulation.

from stratification can also produce localised variations in the nature of the DAS
(e.g., Spica et al., 2020). With poor coupling the observed signal is attenuated,
but will tend to change shape under the influence of structure.
A suitable description of a local source will normally have a strong vertical
component, but a jump or a sledgehammer blow will commonly have some
horizontal effect as well. Such sources can be modelled with a distribution of
surface forces. Active sources such as an explosive shot can be handled with a
moment tensor distribution at appropriate depth.
Local sources
For tap tests, the source point will normally lie close to the cable, but not
right on top. For the closest DAS channels the distances to the ends of the
gauge length and the rotations of the ground velocity to lie along the cable vary
52 Near field

rapidly (Figure 3.1). In favourable circumstances, the DAS recordings show the
equivalent of a refraction spread that can be used to extract information on near
surface structure (as in Figure 5.7).

Source close to cable

As an example of moderate distance propagation in an urban environment
Figure 5.7(a) shows DAS results from a single jump on the sidewalk of a
street in Bern, Switzerland recorded on an optical fibre connecting educational
institutions about 2 km apart (Fichtner, Bowden and Smolinski - private
communication). The cable lies at a depth of 0.7 m, with relatively good
coupling to the surroundings. The gauge length employed was 10 m, and DAS
channels were extracted at 2 m intervals. In this case the jump energy has
propagated to considerable distance and there is noticeable energy in refracted
P waves.
The street lies on glacial deposits with modest variation in properties, but
the situation is not 1-D because of the varying below-ground streetscape with
cellars that can act as side reflectors, as noted above. The data segment in Figure
5.7 is taken from a straight section of the street, but we see that the pattern of
arrivals is not symmetric about the closest channel. This is likely due to side
reflection effects.
It is possible to find a stratified model that provides a good representation
of the major body wave arrivals at a number of different jump points, whilst
being consistent with dispersion results from ambient noise correlation. The
model is shown in Figure 5.7(c) and the simulated DAS strain-rate records
are illustrated in Figure 5.7(b) using a strong vertical force and much weaker
horizontal forces to describe the effect of a person jumping. A separation of 2 m
between source and DAS fibre has been employed. The simulated results show
variations in amplitude and polarity close to the source as predicted from the
strain associated with a dominantly vertical force (Kennett et al., 2024). The 1-D
model cannot allow for any 3-D side effects that can modify the local behaviour.
A consequence of the ground velocity differencing at the ends of the gauge
length is that a single DAS trace can mix aspects of the wavefield with different
character, e.g., refracted and reflected segments with similar timing at the two
ends of the gauge length. Zhu et al. (2021) have reported such issues in a
comparison of a spread geophone and DAS records close to a hammer blow.
The travel time curves superimposed on the DAS records in Figure 5.7 include
the effect of the offset of source from the cable, but not the influence of gauge
length averaging. This averaging leads to some apparently non-causal effects
near the central trace.

5.4 Comparison of in-line and offset sources

As an illustration of both in-line and oblique propagation of seismic waves we
consider two explosive shots recorded on a 3000 m DAS cable at the EastGRIP
5.4 Comparison of in-line and offset sources 53

Shot 22

Shot 1
DAS cable

Figure 5.8 Configuration of the explosive shots illustrated in Figure 5.9 relative to the
3000 m active portion of the DAS cable.

Figure 5.9 DAS strain-rate records for explosive shots at the EastGRIP site in Greenland
from Silixia iDAS with 1000 Hz sampling, and 10 m gauge length: (a) Shot 1 - nearly in
line. (b) Shot 22 at 1050 m from the cable at 2170 m.

site on the North East Greenland Ice Stream. The cable lies at the edge of the
ice stream in an area with relatively uniform ice properties. Each shot uses 200 g
of PETN detonating cord, in the form of a 20 m length doubled to 10 m on the
surface. The detonation velocity is 8400 m/s, so it takes 1.1 ms to detonate a 10
m length.
The configuration of the two shots is shown in Figure 5.8. Shot 1 was placed
54 Near field

100.0
(a) Shot1 (b) Model KS02 ρ [Mg/m 3 ] β,α [km/s] Qβ-1 , Qα-1
90.0 0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50 0.00 0.02 0.04
0.00 0.00

80.0
R5 0.01 0.01

70.0 0.02 0.02

P1
60.0 0.03 0.03
Frequency [Hz]

R2
0.04 0.04
50.0

Depth [km]
R3

Depth [km]
0.05 0.05
40.0
R0 0.06 0.06
30.0 P0
0.07 0.07

20.0
0.08 0.08

10.0 ρ β α
0.09 0.09

0.0 0.10 0.10

0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00
Phase speed [km/s]

Figure 5.10 (a) Frequency – phase speed spectrum from Shot 1, with the control points
used for matching spectra marked. (b) Derived seismic wavespeed and attenuation
model KS02

close to the eastern end of the cable, about 100 m from the cable line. As a result
propagation of seismic waves along the cable is close to in-line, and orientation
effects are minimal. Shot 22 is one of a sequence of shots on a perpendicular
line to the cable about 2150 m along. The shot lies at 1050 m from the cable line
and so waves impinging on the cable span a broad range of angles (see Figure
5.12).
The DAS records for Shot 1 (Figure 5.8 – upper panel) have strong common
mode noise across all channels just at detonation time, most likely due to
pressure effects directly on the DAS interrogator. A prominent air wave from
the surface shot is visible along the full length of the cable, with a propagation
speed of 319 m/s appropriate to the -20◦ C air temperature. The strongest
seismic waves consist of multiply reflected S waves that grade into slightly lower
frequency Rayleigh waves with moderate dispersion. The fundamental mode is
visible out to about 1500 m from the source. Ahead of the S waves is a similar
complex train of multiple P waves with lower amplitude.
The DAS observations for Shot 22 show a well-developed set of Rayleigh
waves followed by a very distinct air wave (Figure 5.9 – lower panel). Ahead
of the main arrivals there are faint but distinct indications of body wave wave
arrivals. The stronger air wave signal on the further end of the DAS cable
suggests that it is more shallowly buried than the nearer end. This effect also
tends to augment the air wave for further distances from Shot 1. The relatively
weak DAS signal for the channels closest to Shot 22 arises from the directional
sensitivity imposed by strain-rate averaging over the gauge length of 10 m.
The clear observations of Shot 1 are very suitable for dispersion analysis. As
in §4.5, Fourier transformation from the time and space domain were used to
extract the frequency – phase speed spectrum (Figure 5.10a). As expected from
5.4 Comparison of in-line and offset sources 55

Figure 5.11 Simulation of DAS strain-rate records for Shot 1 using model KS02: gauge
length 10 m, band pass 1–50 Hz, 0.0025 s sampling, 2500 slownesses out to air wave.

ray–mode duality the strong S wave multiples manifest as well excited higher
modes of Rayleigh waves, and we also see clear pseudo-acoustic P-modes. With
the broad span of frequencies the dispersion of the various mode branches
can be inverted to extract a seismic wavespeed profile with depth with also an
estimate of the attenuation profile. The starting point was a simple compaction
model for the snow of the firn, this was followed by Backus-Gilbert inversion
of Rayleigh mode inversion (Kaplunov et al., 2024) and then direct comparison
of frequency – phase speed spectra was employed to control the P structure
at depth and refine the S wavespeeds. The resulting wavespeed model KS02
is displayed in Figure 5.10b. This model is employed for simulations below to
investigate the specific character of the wavefield.

5.4.1 In-line propagation

We now use the KS02 model to simulate DAS records for Shot 1 by integration
over the frequency-slowness domain using the reflection response of the
medium (Figure 5.11), with a gauge length of 10 m. The source model for the
shot placed at 1 m depth uses a moment tensor formulation M xx = M y y = 1.0,
M zz = 0.5, and all other moment components set to zero. The reduction of the
vertical component allows for the blow out from the surface detonation (seen
on a video of the shot). The cable is assumed placed at the surface with 147
sample points at 20 m intervals.
The DAS strain-rate records are computed with 4096 time points at 0.0025 s
spacing, so a record length of 10.24 s. The band pass is from 1 – 50 Hz, with
tapers from 0.25 – 1 Hz and 50 – 70 Hz. Slowness integration with 2500 values
from 0 to 3.135 s/km is employed with adaptive spacing to improve the accuracy
of the integration. The upper limit of the slowness range corresponds to 318.9
m/s, which is very close to the sound speed at -20◦ C. This means that the
56 Near field

1500 1500 1500

450 m 1050 m
1000 1000 1000

500 500 500

Distance [m]

Distance [m]

Distance [m]
450 m 1050 m
0 0 0

-500 -500 -500

-1000 -1000 -1000

R T R T
-1500 -1500 -1500
Orientation factors Orientation factors

Figure 5.12 Orientation factors for radial (R) and transverse (T) components of strain
rate for sources separated from the cable by 450 and 1050 m. The bland zone with weak
signature for the R component extends as the source moves further away from the cable.

contribution from numerical truncation of the slowness range has the timing
of the air wave, but is not a direct simulation.
The calculated DAS field for model KS02 gives good representation of the
features seen in the observations (Figure 5.9). The timing of the main arrivals
is well reproduced, and the general character of the records is captured. Much
of the visible energy is comprised of higher Rayleigh modes that can also be
thought of in terms of multiple reflections of S waves beyond the critical angle,
with complete reflection at the surface. The explicit contribution from the
fundamental mode Rayleigh wave is weak and is barely visible beyond 1500
m from the source. Direct simulation of the shot is limited by knowledge
of the appropriate source mechanism and source-time function, as well as
the nature of the coupling of the DAS cable to the near-surface firn layer.
A bandpassed delta-function in time and a plausible moment tensor give a
reasonable rendering of behaviour.

5.4.2 Offset shots

The discussion of orientation factors for strain-rate DAS recording in §3.3.3 is
based on a far source, but still provides a good guide to behaviour for sources
that are not too close to the cable. In Figure 5.12 we show the effect of
orientation for sources displaced by 450 m and 1050 m from a 3000 m cable. The
R component, along the cable, relevant to P-SV and Rayleigh waves is shown in
red, and the transverse component T in green.
A consequence of the orientation factor (3.3.9) is that waves impinging at near
normal incidence have very low amplitude in DAS records. This means that for
the section of the cable closest to an offset source there will be a rather bland
zone with low amplitude arrivals extending either side of the closest approach
of the cable to the source. As can be seen in Figure 5.12 the size of this zone
expands for more distant sources because a more limited range of angles arrives
at the same section of cable. The span of distances for which the orientation
5.4 Comparison of in-line and offset sources 57

Figure 5.13 Simulation of DAS strain-rate records for offset Shot 22 for model KS02:
gauge length 10 m, band pass 1–50 Hz, 0.0025 s sampling, 2500 slownesses out to air
wave.

is close to normal increases with source offset. The gradient in amplitude

becomes weaker with a more distant source and so the orientation effect is less
obvious.
A comparable problem of reduced response to P waves at the closest
distances occurs for sources lying beneath a DAS array. With strain-rate
recording the P waves are subdued for steep paths with small horizontal
slowness p, (e.g., Hudson et al., 2021). The effect is reduced but not eliminated
with displacement rate DAS systems where the angular response goes as cos ψ
rather than cos2 ψ. Sensitivity to P waves can be improved by special cable
constructions with engineered fibres, e.g, the helical winding analysed by
Kushnikov (2016), Wuestefeld and Wilks (2019).
The influence of orientation on the recorded wave field appears clearly in
the observed records for Shot 22 (Figure 5.9) and modelling using model KS02
(Figure 5.13). The simulation uses the same parameters as for Shot 1 with
simply modified geometry for the effects of gauge length averaging. Once again
the modelled DAS records provide a good representation of the observations,
notably in the width of the zone of reduced amplitudes.
The sensitivity of the DAS strain-rate field to the relative position of a source
can be readily exploited to estimate source location, particularly with a cable
layout with arms in different directions. Rapid semi-quantitative analysis can be
achieved with arrival time templates including orientation effects, as illustrated
in Figure 5.14. Here the arrival times for P, S and air waves are combined with
traces for phase speeds of 1750, 1500, 1250 and 1000 m/s to provide a rapid
summary of wavefield characteristics. The time traces are modulated in plotting
with the orientation factors for the separation of the source from the cable.
As can be seen from Figure 5.14 where the time templates are compared to
simulations using wavespeed model KS02, this simple scheme works well to
summarise DAS wavefield behaviour. It has proved very useful in disentangling
58 Near field

Figure 5.14 Templates for offset shots including arrival times for different phase speeds
and orientation effects compared with modelling for a surface force source.

overlapping natural events due to surface snow collapse near the DAS cable at
EastGRIP.

5.4.3 Source effects

In the cases discussed above we have used simple source configurations,
but modelling can be accomplished with more complex sources. As with
seismometer records different aspects of the source are reflected in the
character of the excited wavefield, so that with a suitable propagation model,
and representation of cable coupling, inversion for source mechanism is
entirely feasible.
Figures 5.15 and 5.16 show calculations of the DAS field from a source
placed 450 m away from a surface cable using individual force or moment
tensor components. The coordinate system has x along the DAS cable and y
perpendicular, z is in the vertical direction. All calculations were carried out
with the same integration settings, 2500 slownesses from 0 to 3.135 s/km, 4096
time points at 0.005 s spacing and a band pass from 1 – 30 Hz with tapers from
0.25 – 1 Hz and 30 – 45 Hz. The model used is GS07 from the EGRIP aircraft
landing (§4.5). The dominant features of the strain-rate records in all cases are
higher mode surface waves, but there are substantial variations in the character
of the DAS response.
Figure 5.15 shows the effects of surface forces (F x , F y , F z ) again for a point
source 450 m away from the DAS cable. Here Rayleigh waves are produced by
F x , F z and Love waves by F y .
Figure 5.16 illustrates the different DAS fields generated by various moment
tensor components for a point source at 1 m depth, 450 m away from the
DAS cable. The diagonal elements (M xx , M y y , M zz ) are plotted with twice the
amplitude of the off-diagonal terms. Rayleigh waves are the main feature for
M xx , M y y , M zz , M xz and Love waves are strong for M x y , M y z .
5.4 Comparison of in-line and offset sources 59

Figure 5.15 DAS strain-rate simulations for surface forces, 450 m from a DAS cable.
Gauge length 10 m.

Figure 5.16 DAS strain-rate simulations for the various components of a Moment
tensor placed at 1 m depth, 450 m from a DAS cable. Gauge length 10 m.
60 Near field

The difference in the impact of the orientation factors for Rayleigh and Love
waves is clearly present in Figures 5.16 and 5.15. No bland zone appears for Love
excitation, just a rapid change of sign around the closest point as expected from
Figure 5.12.
6
Regional wavefield

In the immediate vicinity of a DAS cable, the representation of the averaged

strain rate as the difference in ground velocity, resolved along the cable, at the
ends of the gauge interval is very effective especially when implemented in the
frequency – slowness domain.
Once a source becomes more distant, the differences in epicentral distance
and azimuth at the ends of gauge length around each DAS channel are small.
In consequence, differencing of ground velocity has the potential for loss of
precision. The problem becomes more acute the larger the distance between
source and cable. Except for very low frequency signals, once the source-cable
separation is more than a few tens of kilometres it is best to use the composite
representation for DAS response (3.3.6) in terms of ground acceleration along
the cable scaled by slowness, with orientation factors for radial and transverse
components of motion. This formulation has the merit that it is insensitive to
the specific mode of strain-rate averaging along the gauge length.
We first consider the near-regional regime out to 500 km or so, where the
wavefield tends to be dominated by multiple crustal reflections of P and S
waves. The influence of the upper mantle becomes more important at greater
distances, with complex returns from the upper mantle transition zone from
15◦ –30◦ epicentral distance.
The presence of slowness scaling in the regional DAS response means that
over the full regional to near-teleseismic distance range the contributions from
seismic phases with steep paths though the mantle are small, and unlikely to
rise above typical noise conditions even with stacking.

6.1 Near-regional regime

As a guide to expected behaviour for DAS recordings at near-regional distances
we compare the radial component as seen by a seismometer with the equivalent
DAS response (Figure 6.1). The simulations use the same flattened version of
the ak135 model, with excitation by a double couple source with M xz , M zx
components placed at 2.5 km depth. The receivers are in line at azimuth 0◦ ,
so that the configuration is optimal for P and SV waves with DAS recording.

61
 62 Regional wavefield

60.0

R
50.0
Time - Distance/8.0 [s]

Sn
40.0

Lg
30.0

20.0

Pg
10.0
Pn

50.0 100.0 150.0 200.0 250.0 300.0 350.0

Distance [km]
60.0

DAS
50.0
Time - Distance/8.0 [s]

Sn
40.0

Lg
30.0

20.0

Pg
10.0
Pn

50.0 100.0 150.0 200.0 250.0 300.0 350.0

Distance [km]

Figure 6.1 Character of near regional seismic wavefield, with comparison of radial
component of seismometer (R) and DAS response for in-line receivers at azimuth 0◦ .
A flattened version of the ak135 model is employed with a M xz double couple source
at 2.5 km depth. A distance amplification factor of X /50 is applied, where X is the
epicentral distance in kilometres. The frequency passband is from 0.03–2 Hz, with a
delta function time source.

Directional influences on DAS are confined to the inclination of the waves in

the vertical plane. The record sections for the seismometer and DAS results are
scaled so that amplitudes are comparable.
As expected for the near-regional environment the wavefield in Figure 6.1 is
dominated by crustal contributions (Kennett, 2002; Chapter 19). There is a very
weak Pn arrival return from the upper mantle with a more significant Pg phase
comprised of crustal multiples. A sequence of multiply reflected S waves in the
6.1 Near-regional regime 63

35 km thick crust merge to form the prominent Lg phase, but again the mantle
arrival Sn is weak.
The modification of the wavefield with DAS strain-rate recording is not large,
but results in a relative reduction of the P wave contributions compared to S at
the larger distances. This arises from the effect of the slowness weighting in the
DAS response (3.3.5).
Figure 6.1 captures the general character of the near-regional wavefield, but
variations in crustal structure and thickness lead to considerable variability in
the times of arrival of the seismic phases. The strong variability in upper crustal
structure affects the timing of Pg and Sg at short ranges. At larger distances the
nature of the interference between multiple crustal reflections forming the Pg
and Lg trains depends strongly on crustal thickness.
We have shown the DAS response over an extended range of distances in
Figure 6.1, but with typical dark fibre a few tens of kilometres long only patches
of the behaviour will be evidenced on a single DAS system. This means that it
can be hard to see the evolution of the complex interference patterns for the
dominant crustal phases across the length of the cable used for DAS recording.

6.1.1 DAS observations

From the comparison above we can expect that near-regional events DAS
recordings will have comparable character to the field as seen by a seismometer.
Nevertheless, the strong orientation effects associated with changes in the
direction of a cable run can modify the appearance.
Figure 6.2 shows DAS recordings, using a colour variable density plot, for
a modest size event (ML 3.0) about 100 km away from a cable in Melbourne,
Australia. The 25 km long ARILT cable is part of a research fibre optic network,
and extends from west to east with segments along or beside roads in the
suburbs of Melbourne. 6252 channels were used at 4 m spacing with a Silixa
iDAS interrogator, measuring strain-rate at 1 kHz with a fixed gauge length of 10
m. Out to channel 2100 the cable azimuth is 95◦ , and after that, toward the east,
there are numerous changes in direction, mostly with orthogonal segments.
The locations of these changes are indicated in Figure 6.2 with black blocks to
the left. In a number of places significant modulation of the DAS response can
be directly related to the changes in cable direction.
This event is too close to the DAS cable for the emergence of any upper mantle
phase, and so the distinctive wavefronts seen in Figure 6.2 correspond to Pg and
Sg. The coda of each of these phases shows a rather complex character, with
multiple scattering associated with the varied streetscape and changing near
surface conditions.
In the second example of a regional event recorded in an urban environment
Figure 6.3 shows two 400 m segments from a 25 km long cable (KNSG) in Perth,
Western Australia. The ML 4.8 event near Cadoux on the Yilgarn craton on 2022
January 24 is about 185 km away from the cable and the wavefronts cross at
64 Regional wavefield

Figure 6.2 Local event recorded on the 25 km ALIRT DAS cable in Melbourne, Australia
with 4 m channel spacing. The M3.0 event (2022 February 26) lies about about 102 km
from the interrogator unit at the west end of the cable, with waves crossing the nearly
east-west oriented cable at azimuths around 317◦ . The traces are bandpass filtered
from 0.5–2 Hz. The positions of major changes in cable orientation from approximately
east-west are marked by black blocks at the left. [Courtesy of Voon Hui Lai & Meghan
Miller].

an azimuth of 233◦ . The crustal thickness on the craton is about 40 km and so

the Pn, Sn phases have just started to emerge from the crustal multiples. These
upper mantle phases are just visible at the front of their respective wavetrains
on the highest amplitude channels.
The DAS cable weaves its way through the streets of Perth with typical
segments with consistent orientation about 100–200 m long followed by a right
angle turn to a new direction. The very strong influence of cable direction is
manifest in the almost total loss of amplitude in the short segments where the
cable line is almost orthogonal to the incoming wavefront. As expected from
the discussion in §3.5, the transition in amplitude extends over several gauge
lengths at each turn in the cable, since cable loops may be introduced as well as
the directional change. Besides the strong changes in amplitude there are also
more subtle dips in amplitude that are likely to be the result of variable cable
coupling to the subsurface.
6.1 Near-regional regime 65

980 3
2
5
1
1
1
1
1
1
1
990 2
2
2
3
Channel number

2
2
2
1
1
5
1000 1
1
2
2
1
8
1
1
1
2
1010 3
3
3
2
3
2
2
3
2
2

20.0 P 40.0 S 60.0 80.0

Time [s]

200

210
Channel number

220

230

20.0 P 40.0 S 60.0 80.0

Time [s]

Figure 6.3 Event near Cadoux, Western Australia (2022 January 24) recorded on the
KNSG cable in Perth, with 10 m channel spacing. The ML 4.8 event is about 185 km away
from the interrogator, with waves crossing the north-south trending cable at azimuths
around 233◦ . Two 400 m segments of the DAS channels are shown that display the
strong effects of cable orientation on signal amplitude, with right angle turns in the
cable run as it follows different streets. The segments with markedly reduced amplitude
lie close to orthogonal to the path to the event, [Courtesy of Erdinc Saygin].
66 Regional wavefield

Apart from the modulation of the waveforms in Figure 6.3 by orientation

and coupling effects, we see significant variations in the character of the Pg,
Lg crustal contributions. Wave packets tend to be coherent across a distance of
100 m or so, but then there can be a change in the prominence of different parts
of the wavetrain. Such effects can arise from scattering along the path from the
source and variations in the local conditions in the near-surface sediments.
The use of a ‘wiggle-trace’ presentation in Figure 6.3 provides more insight
into the detailed properties of the regional phases than the variable area style
used in Figure 6.3. The variable area form is good for visual phase detection
and assessment of amplitude effects, but less helpful for the nature of the
waveforms.

6.1.2 Simulating the regional DAS field

The previous two examples show the potential complexity of the regional
wavefield, especially when recorded on urban dark fibre with short orthogonal
sections. I now show how it is possible to make a more quantitative approach
to the assessment of regional events, even in regions without dense seismic
coverage.
Figure 6.4(a) shows the recording of an earthquake near Jindabyne in the
Snowy Mountains of New South Wales on a single-mode DAS cable at the
Tidbinbilla Deep Space Tracking Station south of Canberra in Australia (Lai,
Miller & McQueen – private communication). A spare fibre leading from the
central building to a temporarily unused large-dish antenna to the north was
used for the DAS system. The cable lies in a conduit at a depth of 2 m below the
uneven surface and its path follows a set of straight-line segments with some
sharp changes of orientation (Figure 6.4b). The surface conditions across the
site vary, with ground disturbance in places associated with past building work.
The site shows considerable topography with some sharp slopes along the cable
run. As mentioned in §5.2, the cable configuration was calibrated by tap tests,
some of which showed significant influence from the surface slope (Figure 5.6).
There are complications in the cable geometry around the 700 m mark where
an excess cable loop has been introduced. This loop probably accounts for the
rapid changes in recorded signal near this location (Figure 6.4a).
The various large antennas at this site are used to track satellites and missions
to the outer planets with frequent movement that generates significant low
frequency noise. As a result the usable frequency band is limited to above 1 Hz.
This means that the DAS data lie in the regime where influences from scattering
are strong, and the coherence of regional signals is reduced.
The Snowy Mountains region shows occasional earthquake activity (e.g.,
Bock & Denham, 1983) with events up to ML 5. The current distribution
of seismic stations provides adequate control on location, but the absence
of any close station means that depth control is limited. The ML 3.2 event
near Jindabyne occurred during the limited period when the DAS system was
 6.1 Near-regional regime 67

(a) DAS Strainrate - observations (b) Cable geometry

800 35.395
S
890 m
750

700
N
748 m
685 m
650
631 m
Distance [m]

600
502 m

550 400 m

351 m

500
278 m

218 m
450
0.00 km 0.25 158 m

35.401
400 148.978 E 148.982

20 P 30 S 40 50
Time [s]

Figure 6.4 Regional event: Jindabyne earthquake ML 3.7 of 2021 May 5 recorded at
the Tidbinbilla Deep Space Tracking Station south of Canberra, Australia. The data are
high-pass filtered above 1 Hz to eliminate strong site noise. (a) strain rate observations
for the DAS cable, filter 1.0-30.0 Hz, with distance scale along the cable in metres; (b)
configuration of the DAS cable. [Courtesy of Voon Hui Lai & Meghan Miller].

operational in 2021. The event lies to the south of the recording site at a distance
of 125 km; waves cross the cable with azimuth 21.5◦ . The recordings, with
a high-pass filter above 1 Hz, display a relatively strong S arrival with some
variation in amplitude and a less pronounced P arrival. The modulation of the
amplitude of the observed arrivals with position along the cable and changes in
the structure of the waveforms can be linked to the varying orientation of the
segments of the DAS cable (Figures 6.4b, 6.6b).
For simulation of these DAS observations we need both a structural model
and a source representation. The moment tensor source was based on previous
larger events in the area, since this event was too small for a direct estimate to
be made with the limited available network. The radiation pattern for P and S
waves for the source used is shown in Figure 6.5(a). The azimuth to the DAS
array lies close to a node in the P wave radiation pattern and the dominant S
contribution comes from SH waves. Both the source and the DAS recording
station lie in a zone with thickened crust, but there are pronounced wavespeed
gradients in the crust (e.g., Salmon et al., 2013) so that it hard to find a suitable
stratified model for high-frequency simulation. A regional velocity model with
68 Regional wavefield

(a)
N N N

P SV SH

E E E

(b)

ρ [Mg/m3] β,α [km/s]

0 1 2 3 4 5 6 7 8 9
0
sm45
5

10

15

20

25
Depth [km]

30

35

40

45

50
ρ β α
55

60

Figure 6.5 Regional event: Jindabyne earthquake ML 3.7 of 2021 May 5:

(a) Radiation pattern for P and S waves with take off for P indicated; (b) sm45 structural
model with 45 km thick crust. The P wavespeed (α) is shown in red and the S wavespeed
(β) in green, with the associated density (ρ) profile in pale blue.

a gradient zone at the base of the crust from 35–45 km has been used (Figure
6.5b). This sm45 model represents a simplified version of P wavespeed profiles
derived from refraction studies in the Snowy Mountains area. The conversion
from P to S wavespeeds is based on regional results.
For this regional distance range we use DAS strain rate simulation based on
the scaled acceleration formulation (3.3.6) in the frequency – slowness domain.
Slowness integration was taken out to 0.33 s/km with 1200 slownesses and a
frequency passband from 0.1 to 8.0 Hz was used. From several trials a source
depth of 6 km was selected as providing the best representation of wavefield
character (Figure 6.6a).
A modest amount of signal generated noise has been added to the synthetic
records, using the approach described by Kennett (1985) in which a random
 6.1 Near-regional regime 69

(a) DAS Strainrate - simulation (b) Orientation factors

800 800
N

750 750

700 700

650 650
Distance [m]

600 600

550 550

500 500

450 450

R T
400 400
20 P 30 S 40 50
Time [s]

Figure 6.6 Regional event: Jindabyne earthquake ML 3.7 of 2021 May 5. (a) DAS
modelling using integration over slowness and frequency with a source set at 6 km
depth; (b) orientation factors for P, SV waves (R) and SH waves (T), with the azimuth of
the arriving wavefront indicated – distance scale along the cable in metres as in (a).

component modifies the frequency-slowness spectrum. The synthesised

regional seismograms are still too simple compared with the DAS observations
in Figure 6.1(a), but the pattern of visibility of the P and S arrivals shows a
reasonable correspondence with the observations. The synthetics only include
propagation along the great-circle path, but because the azimuth to the DAS
system is close to a P wave radiation node, scattering effects from off path have a
significant influence (as seen in the observations). The relative amplitude of the
S waves are enhanced by both the scaling by horizontal slowness from (3.3.6)
and the radiation pattern effects. The orientation effects for the radial and
tangential components of strain produced by the cable geometry are displayed
in Figure 6.6(b) alongside the synthetics on the same distance scale, and tie well
with the distance variations in the synthetics.
Comparison of the simulations with the observations reveals a zone of
reduced S wave amplitude between cable distances of 550 and 650 m where the
orientation factors are favourable for SH waves. This amplitude reduction can
be linked to a change in the ground conditions in the material surrounding the
conduit where it parallels a road. Such an effect could not be recognized without
the aid of the modelling.
70 Regional wavefield

6.1.3 DAS studies of regional seismic sources

The dense sampling afforded by DAS records has recently been used at regional
ranges in California to improve focal mechanism solutions and to look at the
details of the rupture processes in sizeable earthquakes (J Li et al., 2023a,b).
J Li et al. (2023a) have demonstrated how polarity results extracted from
DAS observations can be used to supplement picks made from seismometer
networks and fill in azimuth gaps to improve the accuracy of focal mechanism
determination and reduce uncertainty. The combination of unidirectional
sensing and relatively low signal strength compared with noise for small local
to regional events, means that direct determination of first-arrival polarities
on DAS is difficult. However, by the use of multi-channel cross-correlations
between earthquake pairs relative polarities for the events can be established
along the channels of the cable. The absolute values are then calibrated
with robust P-polarity measurements from the vertical component of nearby
broad-band stations. Changes in P-polarity along a segment of DAS cable can
be particularly valuable in constraining nodal lines and hence improving focal
mechanism solutions.
A more direct use of DAS observations for regional events is presented by J Li
et al. (2023b), who have used a 100 km long fibre-optic cable as a dense DAS
local array for imaging the source processes of a Mw 6.0 crustal earthquake
near the California-Nevada border. The DAS cable stretches in a direct line
away from the event for most of its length, so that the regional wavefield is
captured in detail over more than 80 km. DAS waveforms of aftershocks are
used as empirical Green’s functions (EGFs) to extract estimates of sub-event
locations using both the P and S wavetrains. Four separate sub-events can be
well constrained, with hints of greater complexity. These sub-events can be
validated by conventional seismometer observations, but these alone are not
sufficient to extract the detailed processes revealed from the DAS results.

6.2 Far-regional to near-teleseismic regime

The transition from the regional range into the teleseismic regime covers the
zone of body-wave interaction with the transition zone in the upper mantle,
that introduces complex triplications for both P and S phases. Once again we
compare the radial component as seen by a seismometer with the equivalent
DAS response (Figure 6.7). The simulations use a flattened version of the ak135
model extending to 1000 km depth. A double couple source with M xz , M zx
components is placed at 5 km depth. The receivers are in line at azimuth 0◦ , so
that the configuration is optimal for P and SV waves with DAS recording. The
record sections for the seismometer and DAS results are again scaled so that
amplitudes are comparable.
In Figure 6.7 the time window is extended to span from the P arrivals to the
Rayleigh surface waves. In this distance range we begin to see the influence of
 6.2 Far-regional to near-teleseismic regime 71

180.0

160.0

140.0

120.0 R
100.0
Time - Distance/8.0 [s]

80.0

60.0

40.0

20.0 S
0.0

-20.0

-40.0
P
-60.0 R
-80.0
500.0 1000.0 1500.0 2000.0
Distance [km]

180.0

160.0

140.0

120.0 R
100.0
Time - Distance/8.0 [s]

80.0

60.0

40.0

20.0 S
0.0

-20.0

-40.0
P
-60.0 DAS
-80.0
500.0 1000.0 1500.0 2000.0
Distance [km]

Figure 6.7 Character of far-regional field, with comparison of radial component of

seismometer (R) and DAS response for in-line receivers at azimuth 0◦ . A flattened
version of the ak135 model to 1000 km depth is employed with a M xz double couple
source at 5 km depth. A distance amplification factor of X /1000 is applied, where X
is the epicentral distance in kilometres. The frequency passband is from 0.01–0.25 Hz,
with a delta function time source.
72 Regional wavefield

weighting by slowness and the enhanced high-frequency response of DAS with

scaled directional acceleration rather than ground velocity as recorded on the
seismometer. For DAS the amplitude of the P arrivals is reduced relative to S, but
the major change comes in the surface wave regime. The low frequency surface
waves are suppressed relative to the higher frequency components. Here we
have used a flat source spectrum from 0.01-0.25 Hz, whereas we would expect
stronger excitation of the lower frequencies for a realistic earthquake source.
Slowness scaling enhances the amplitude of the surface waves so that in-line
DAS recordings tend to be dominated by the surface wave components – this
can be seen in the shorter range records in Figure 7.5.
The evolution of the wavefield with distance is still quite strong in this
far-regional range, but near broadside observations on DAS cables will be
coherent over a fair length of cable especially for moderate frequencies (below
1 Hz). As a result stacking of a bundle of DAS channels can be effective for
enhancing signal relative to noise. The stacking process can also be thought of
as expanding the effective gauge length and so allowing the resolution of slower
variations in strain rate.
7
Global wavefield

Just as for an individual seismometer, a length of DAS cable only samples

a very small fraction of the global seismic wavefield. Yet, it is important to
understand the nature of the DAS signal as a function of epicentral distance so
that observations can be put into context.

7.1 The influence of DAS recording

As we have seen in §3.3 the DAS field for a source distant from the cable
can be represented as slowness weighted ground acceleration modulated by
orientation factors. Thus DAS strain rate recording favours energy travelling
near horizontally, and so strong medium frequency Rayleigh waves are a
common part of DAS records. Love waves are less frequently observed because
they are best expressed when they cross the DAS cable at 45◦ . Mantle P
phases can be rendered quite well by DAS for strong enough events, but phases
penetrating the core tend to be suppressed. In principle, S mantle phases
would be enhanced, but the orientation factors for SH waves are generally
unfavourable and so this aspect of the wavefield is not well represented.
The influence of wave slowness on the character on the observable DAS
seismic wavefield is shown in Figure 7.1. The familiar plot of the travel times
of major seismic phases is here modulated by slowness. Full strength colour
is applied for a slowness of 25 s/◦ appropriate to multiply reflected S waves,
and reduced to feint for nearly vertically travelling waves. In this way a visual
impression is provided for the aspects of the wavefield that are likely to be
important on any particular DAS cable.
The slowness modulation associated with strain tends to enhance the
significance of P and S multiples relative to the direct arrivals. The result is that
the wavetrains for P or S tend to increase in amplitude with time as the slowness
effects of strain compensate for the normal decrease seen on seismometer
traces (cf. Figures 7.3, 7.4).
In Figure 7.1 we also show the time bands for the arrival of longer period
fundamental mode surface waves (frequency < 0.03 Hz) for both Rayleigh waves
R and Love waves L. The Love waves emerge from the back of the sequence of

73
74 Global wavefield

50.

R L

PKPSKP

PKPPKP
40.

P3KP

ScSScS PKKP
30. SSS PcPPKP
SKKS
Time [min]

SP
SS
PPP
SKS
SKiKS SKP
SKiKP
20.
PKP
PcPPcP PKiKP
ScS
S
PP
ScP

10.
PcP
P

0.
0. 20. 40. 60. 80. 100. 120. 140. 160. 180.
Delta [deg]

Figure 7.1 Travel times for major seismic phases: ak135 model with surface source.
Phases arriving as P waves are shown in red, and those arriving as S in green. Slowness
weighting along each phase branch is applied, with the colour density increasing with
larger slowness. The arrival bands for lower frequency Rayleigh (R) and Love (L) waves
are also indicated.

multiple SH mantle reflections that can also be described as higher modes. The
strongest component of Rayleigh waves on continental paths is the Airy phase
at around 0.05 Hz and this lies just behind the marked R band.
7.2 Teleseismic observations 75

Figure 7.2 Mw 7.3 event to NE of Timor-Leste on 2021 December 29, depth 165
km, recorded on the western end of the ALIRT cable in Melbourne, Australia with an
epicentral distance of 34.1◦ to the DAS cable segment displayed. The S wavetrain is
displayed, P is near nodal. The records are filtered with a pass band from 0.02–0.1 Hz.
The location of major changes in cable direction are marked by black blocks to the left.
[Courtesy of Voon Hui Lai & Meghan Miller].

If we look at any particular distance we see a wide range of slowness

associated with different phases. Except where phases cross and interfere,
conversion from a stack of nearby DAS channels to equivalent ground velocity
will generally be feasible for a small segment of record with consistent
properties. There is no general conversion factor, so that it is important to
understand and analyse the DAS wavefield in its own right.

7.2 Teleseismic observations

Figure 7.2 displays DAS records from the ALIRT cable in Melbourne, Australia
for an intermediate depth earthquake to the northeast of Timor-Leste in the
Banda arc. The event lies 34.1◦ away from the recording zone, and so just
into the teleseismic range. A Silixa iDAS interrogator was employed, measuring
strain-rate at 1 kHz with a fixed gauge length of 10 m; the channel spacing was
4 m. A colour variable intensity display is presented here for the S wavetrain,
with distinct S and sS arrivals followed by a strong SS train. The S wavefronts
cross the cable at about 42◦ inclination and so the situation is very favourable
76 Global wavefield

Figure 7.3 Teleseismic event waveforms with dominant frequency range 0.01–0.05
Hz normalized by event (black: DAS, strain cable azimuth 142 ◦ ; grey: seismometer,
velocity oriented along cable). Upper traces – Honduras event Mw 7.5 of 2018 January
10, epicentral distance 39.1◦ , azimuth 130◦ ; lower traces – Mexico event Mw 7.2 of
2018 February 16, epicentral distance 30.2◦ , azimuth 139◦ . [Adapted from Lindsey et
al. (2020)].

for both polarisations. In contrast, the P waves are near nodal and so poorly
expressed. A weak ScP phase is also visible, just ahead of sS.
The cable segment displayed in Figure 7.2 has many changes in cable
direction and the interruptions in the character of the wavetrain correlate quite
well with the influence of variations in the orientation factors. The presence
of both SV and SH contributions means that there is not a simple mapping
even with orthogonal cable segments. Stronger ScP is associated with dominant
SV waves. The segment at the far end of the cable is noisier and there are also
suggestions of variation in cable coupling in this zone.
As a comparison of the differences between events as recorded on a
seismometer and DAS records, Figures 7.3, 7.4 display four events from an
experiment near Sacramento, California (Lindsey et al., 2020). A long straight
section of DAS cable runs at azimuth 142◦ and the broad band seismometer lies
about 66 m away from the cable. The horizontal ground velocity records from
the seismometer are rotated to lie along the direction of the cable to provide the
closest tie to the unidirectional strain records from the DAS system. The DAS
records were stacked over five gauge lengths (50 m) from the channels closest
to the seismometer location.
The events in Figure 7.3 are from the early teleseismic range (30–40◦ ) with a
great-circle paths from the source making a modest angle to the cable direction.
7.2 Teleseismic observations 77

Figure 7.4 Teleseismic event waveforms with dominant frequency range 0.01–0.05
Hz normalized by event (black: DAS, strain cable azimuth 142 ◦ ; gray: seismometer,
velocity oriented along cable). Upper traces – Papua New Guinea event Mw 7.5 of 2018
February 25, epicentral distance 97.1◦ , azimuth 231◦ ; lower traces – Peru event Mw 7.1
of 2018 January 14, epicentral distance 63.7◦ , azimuth 141◦ [Adapted from Lindsey et al.
(2020)].

For the closer Mw 7.2 event in Mexico (30.2◦ ) the strong P wavetrain on the DAS
record extends over and obscures the S arrival that can be discerned on the
seismometer trace. The onset of P is also more distinct on the seismometer.
The slightly more distant event from Honduras (39.1◦ ) arrives at a relatively
unfavourable angle for the presence of Love waves, but these are strong enough
to appear on both the seismometer and DAS records. The initial P arrival is weak
and barely detectable on the DAS trace, where slowness weighting enhances the
visibility of PP. The later portion of the S wavetrain is also more prominent than
on the seismometer record. The complex and elongated suite of Rayleigh waves
suggest the presence of multipathing between the source and the receivers.
The events shown in Figure 7.4 lie at much greater distance. For the Peruvian
event at 63.7◦ epicentral distance the great-circle path lies almost in line with the
cable and so is optimally oriented in the horizontal plane for P and SV waves.
The slightly noisy conditions and the steeper vertical incidence of P than the
other phases means that it cannot be distinguished on the DAS trace, though
present on the seismometer. Both S and SKS are seen, even though the core
phase has smaller slowness. Once again the Rayleigh waves are complex and
of long duration. The Papua New Guinea event at 97.1◦ epicentral distance
lies just at the limit of the mantle body wave branches where they merge with
78 Global wavefield

Figure 7.5 Türkiye earthquake Mw 7.8 of 2023 February 6: Record section of strain-rate
records from contributors to the Global DAS month plotted against epicentral distance.
Traces are filtered between 0.1 and 3 Hz and normalized to the maximum absolute value
of the first 2400 s after origin time. The traces are superimposed on a slowness-weighted
travel time plot as in Figure 7.1. [Modified from Wuestefeld et al. (2023)]

core reflections. Despite a noisy background, the DAS record shows a number
of distinct body wave phases. The surface waves arrive obliquely on the DAS
cable after a long passage on paths across the oceans and display the style of
7.3 Simulating the teleseismic DAS field 79

dispersion characteristic of such paths with little difference in group velocity

for frequencies below 0.02 Hz.
In February 2023 a major international effort was made to coordinate DAS
recording across the globe with 32 DAS systems acting together as a monitoring
network in the Global DAS month. One hour segments of data for events above
magnitude 5 reported by the US Geological Survey were shared by participants
(Wuestefeld et al., 2023). Many of the DAS systems in the northern hemisphere
were well placed to record the Mw 7.8 event in Türkiye, a complex event with
successive failure on multiple fault segments that caused extensive damage.
This large event was recorded to beyond 100◦ epicentral distance (Figure 7.5).
This cooperative international venture provides the first direct sampling of the
global DAS wavefield at many stations. In Figure 7.5 the record section of
DAS traces from the different sites has been superimposed on a suitably scaled
version of the slowness weighted travel time plot presented in Figure 7.1. This
record section concentrates on higher frequencies and so the dominant energy
of Rayleigh waves lies just behind the R band marked for lower frequencies. The
phase–time underlay provides an effective summary of the likely observable
phases on the DAS traces. Phases with poor slowness weighting are simply
not seen in Figure 7.5. The under-representation of S and S multiples can be
ascribed to orientation issues that are more important than for P phases.

7.3 Simulating the teleseismic DAS field

As we have seen for events at teleseismic distances the dominant contribution
on DAS records at low frequencies will be surface waves, usually dominantly
Rayleigh waves, enhanced by the amplification of acceleration by horizontal
slowness. Under favourable circumstances it can be possible to pick up
teleseismic body-wave arrivals on dark fibre in urban areas (e.g., Figure 7.2),
though often high-pass filtering is needed to suppress noise.

7.3.1 Body waves

Figure 7.6(a) displays the DAS recording of the Mw 8.1 earthquake in the
Kermadecs (2021 March 04, 29.723◦ S, 177.279◦ W, 30 km depth) on an urban
cable in Perth, Western Australia connecting CSIRO establishments (Saygin and
Ricard - private communication). The cable run follows the pattern of streets
and includes many rapid changes in direction (Figure 7.6b). The event lies at
56.5◦ away from the cable, with a takeoff azimuth at the source of 249.4◦ (Figure
7.7). The wavefield crosses the DAS array at an azimuth of 286.6◦ , so that some
parts of the cable run are particularly well placed to record P waves, while others
are almost orthogonal. A consequence of these changes in orientation of the
cable relative to the arrival direction from the source is that there is distinct
banding in the amplitude of P in the DAS records. In addition, Figure 7.6(a)
shows some amplitude variations that can be linked to changing coupling of
the DAS cable to the ground (e.g., in the street segment from 1.00–1.25 km).
 80 Global wavefield

(
a) DASs
tra
inr
ate-o
bse
rva
tio
ns (
b) Ca
blege
ome
try

PP
P S

Figure 7.6 Teleseismic event: Kermadec earthquake Mw 8.1 of 2021 March 4 recorded
on the southern portion of the KNSG urban cable in Perth, Western Australia. (a) Strain
rate observations for the DAS cable, filter 0.1-1.0 Hz; (b) configuration of the illustrated
portion of the cable, superimposed on Open Street Map. [Courtesy of Erdinc Saygin].

For a scenario like this where a few phases can be recognised, it is most
effective to use modelling procedures that are specifically oriented towards the
representation of specific phases (see, e.g., §16.1.2 of Kennett, 2001). For the
main teleseismic body waves for shallow sources it is possible to use a simple
representation for propagation in the mantle but provide a more complete
representation of the crustal response at both source and receiver sides (e.g.,
Marson-Pidgeon & Kennett, 2000). This scheme is well suited to adaptation
to DAS since it employs an integration over slowness and frequency for a
bundle of slownesses around the ray theoretical value for the source-receiver
separation, and can be used directly with slowness weighted acceleration
modulated orientation factors appropriate to the receiver locations (3.3.6). The
advantage of using a bundle of slownesses is that the full interactions of the
direct phases P, S with their associated depth phases are represented.
For the Kermadec event, the ak135 model (Kennett et al., 1995) was used
for both source and receiver crustal structure with mild attenuation included
[Q p−1 = 0.001, Q s−1 = 0.002]. For this event in the medium teleseismic range it is
sufficient to use ray theory for time delay and attenuation. The mantle model
was again ak135 with the loss factors increased to Q p−1 = 0.005, Q s−1 = 0.001
between 120 and 660 km depth.
For the source the CMT solution was used with a point source at 29 km depth.
The expected radiation patterns for this event are displayed in Figure 7.7, with
the takeoff angle for the P waves indicated. The simulation used a trapezoidal
 7.3 Simulating the teleseismic DAS field 81

N N N

P SV SH

E E E

Figure 7.7 Teleseismic event: Kermadec earthquake Mw 8.1 of 2021 March 4: radiation
patterns for P and S waves, with takeoff for P waves indicated.

(
a) DASs
tra
inr
ates
imul
ati
on (
b) Or
ient
ati
onf
act
ors

P S

Figure 7.8 Teleseismic event: Kermadec earthquake Mw 8.1 of 2021 March 4 recorded
on an urban cable in Perth, Western Australia. (a) DAS strain-rate simulation using
a slowness bundle around the geometric slowness, the S portion is shown at higher
amplification for clarity; (b) orientation factors for P, SV waves (R) and SH waves (T),
with the azimuth of the arriving wavefront indicated – distance scale in kilometres as in
(a).

wavefunction 5 s long with rise and decay times of 1.25 s. The frequency band
employed was from 0.05–1 Hz, and a bundle of 120 slownesses around the
geometrical ray value was used separately for the P and S arrivals.
Because the recording channels along a DAS fibre are closely spaced the
geometric slowness varies slowly with position along the cable. As a result,
the same bundle of slownesses can be employed in modelling for a sweep of
channels, thereby reducing computational effort significantly.
The simulated DAS records in Figure 7.8(a) reproduce well the modulation of
the P arrivals by orientation. There is a hint of the shift in relative amplitude
between the first arriving direct phases and the following depth phases seen
82 Global wavefield

in the observations (Figure 7.6a). The complex pattern of orientation effects for
both radial and transverse motion due to the changes in the DAS cable direction
are shown in Figure 7.8(b) on the same distance scale as the synthetics and
observations. Some slight shifts in the position of amplitude minima between
the observations and the modelled results are likely to be a product of an
imperfect mapping of channel position due to the presence of cable loops.
Both the simulation and the DAS observations show reasonable consistency
across the array in the style of waveforms, even though the amplitudes are
affected by orientation. It is therefore interesting to see what can be achieved
with simple stacking of the modelled DAS traces. Stacks of the P and S
wave traces, normalized by the number of traces, are shown at the top of the
simulation panel (Figure 7.8a).
For the P waves, the stack trace provides a good rendering of the character
of the arrivals with little distortion, even though there are notable variations
in amplitude. However, for S there are significant variations in polarity of
the contribution from the SH waves on the transverse component to the path
from the source (Figure 7.8b). The polarity changes are most noticeable at the
beginning of the S packet in the DAS simulations. Their net effect on stacking is
to almost entirely suppress the onset of S in the stack trace leaving only modest
depth phase contributions from the radial component (SV ).
Although the simulated records generally show a strong similarity with the
observations, there are instances of notable discrepancy. In particular, the
significant amplitude difference between the modelling and the observations
around 1.25 km indicates a cable segment along a road with very different
coupling conditions from the rest of the cable. Although the synthetics in
Figure 7.8(a) indicate little change in P waveform along the cable, there are
some distinct shifts in the pattern in the observations consistent with a higher
emphasis on the depth phases for some parts of the cable (Figure 7.6a).
Slowness changes for the main path are too small to make any significant
difference, so such effects likely arise from local variations in near surface
structure along the cable run.
The observations in Figure 7.6(a) show strong coherence of the P arrivals over
nearly 400 m for road segments with consistent direction. Changes in coherency
between different road segments could be associated with the nature of the
cable conduit and surrounding conditions.
The success of this relatively simple modification of the teleseismic response
with the slowness and angular weight to simulate DAS records suggests that
waveform inversion for source characterisation will be feasible. With stacking
of a number of DAS channels on a well oriented cable segment, to increase
the signal to noise ratio, the DAS results could be used directly in an inversion
scheme, such as that described by Kennett et al. (2000), along with seismometer
records.
7.3 Simulating the teleseismic DAS field 83

7.3.2 Surface waves

The approximation of the DAS response with slowness and orientation factors
applied to acceleration (3.3.6) is suitable for both regional and teleseismic
distances, For surface wave synthesis using modal summation, the DAS form
can be included by simple modifications of the receiver terms for each mode
before frequency integration. For a mode with frequency ω and angular order l
the slowness p = ω/(l + 12 ).
8
Ambient noise correlation

The exploitation of correlations between seismic traces has opened up new

ways of studying the Earth, since the correlation wavefield emphasises different
aspects of seismic wave propagation than are revealed by direct excitation
by a source. The process of correlation can be interpreted as exploiting
constructive interference between portions of seismogram time series. This
interference enhances similarities in the traces that are separated in time, and
thereby brings to light features that individually have small amplitude but
collectively become significant. The correlation wavefield associated with a
pair of stations resembles aspects of a virtual source-receiver pair, so that the
extracted component can be interpreted in terms of subsurface structure.
In the context of ambient noise, primary attention is focussed on waves
with near-horizontal propagation, such as microseisms or urban noise. In
consequence, surface-wave contributions emerge most readily from stacked
cross-correlations between stations. Successful results are often achieved even
though the situation does not meet theoretical expectations (see, e.g., Nakata et
al., 2019).
In the ideal conditions of a uniform distribution of uncorrelated noise
sources, the cross-correlation of seismic records between two stations is closely
related to the Green’s function for the path between them (Fichtner & Tsai,
2019). A number of different derivations have been made with different
assumptions such as a diffuse wavefield (Lobkis & Weaver, 2001), energy
equipartitioned among surface wave modes (Weaver, 2010), or with sources
on a boundary surrounding the two stations (Wapenaar, 2004; Wapenaar &
Fokkema, 2006). However, a typical noise distribution is far from uniform, and
so we here start directly with the cross-correlation of seismic records between
two receiver locations. We show how the properties of local correlation can
be understood as arising from the interaction of the influence of distributed
sources, and examine how far the result can approach an approximation to the
Green’s function.
When the two receiver elements are part of a DAS system, their geometric
relationship is defined by the line of the fibre-optic cable and there are strong

84
8.1 Strain correlation 85

directivity effects. When cable is specifically deployed for an experiment, it may

be possible to gear the configuration to the directivity of likely noise sources.
However, for dark-fibre exploiting existing telecommunication channels the
cable geometry is already imposed and can have a significant influence on the
nature of possible correlations. Nevertheless, useful results can be achieved in
circumstances that may seem unhelpful, such as a DAS cable running along a
major highway (Yang Y. et al., 2022).
There have been many successful applications of ambient-noise
interferometry for DAS data for imaging the structure in the near surface.
The most common approach has been to use interferometry between DAS
channels on the same linear segment of cable to extract Rayleigh wave
dispersion from noise cross-correlation functions (e.g., Dou et al., 2017; Martin
et al., 2017; Zeng et al., 2017; Ajo-Franklin et al., 2019; Yang et al., 2022; Kennett
et al, 2023). With channels from cable segments with different orientation there
can be mixing of Rayleigh and Love mode contributions in the correlations (e.g.,
Luo et al., 2020). Ji et al. (2024) have demonstrated the successful extraction of
Love waves from parallel but displaced cable segments.
We here build on the analysis by Kennett et al. (2023) to look at the nature
of DAS correlation and sensitivity to noise directivity. For simplicity we use
a structure that depends solely on depth and so can represent the seismic
wavefield at each location in frequency-slowness space, as in the earlier
discussions.

8.1 Strain correlation

We adapt the theoretical development from Kennett et al. (2023) for
inter-receiver correlation of strain. For simplicity we concentrate on a single
frequency ω and consider the a strain component from an isotropic source with
source spectrum M (ω). At each receiver location we include an orientation
factor h i (ω) to allow for the directional effects of strain recording.
We use a frequency – slowness representation, as in §3.2, so that for a receiver
a distance X from a source we can represent the resulting strain record in the
frequency domain as an integral over slowness p of the response of the local
stratified medium G e (p, ω) multiplied by a horizontal phase term:
h ω i1/2 Z p0
e(X , ω) = M (ω)h(ω) dp p 1/2G e (p, ω) exp[iωp X ]. (8.1.1)
X 0

Here we have assumed that we can use the high frequency asymptotic form for
the horizontal phase dependence. For the same source recorded at receivers 1
and 2 at distances X 1 , X 2 the cross-correlation E 12 (ω) is just a multiplication in
the frequency domain and so

E 12 (ω) = e(X 1 , ω)e(X 2 , ω)∗ . (8.1.2)

The cross-correlation therefore depends on the difference of the phase of the

86 Ambient noise correlation

contributions from the two receivers. This is the property that allows the
emergence of path-related effects in the presence of many sources.
Using the representation (8.1.1) the cross-correlation can be written as
ωh 1 (ω)h 2 (ω)
Z p0
12 2
E (ω) = |M (ω)| dp p 1/2G e (p, ω) exp[iωp X 1 ]
(X 1 X 2 )1/2 0
Z p0
1/2 ∗
· dq q G e (q, ω) exp[−iωq X 2 ]. (8.1.3)
0

We now introduce the distance between the two stations X 12 = |X 1 − X 2 | −

δX 12 and recast the second slowness integration in terms of the difference in
slowness ζ = p − q. Then
ωh 1 (ω)h 2 (ω)
Z p0
E 12 (ω) = |M (ω)| 2
dp p 1/2 exp[iωp(X 12 + δX 12 )]
(X 1 X 2 )1/2 0
Z p0
· dζ (p − ζ)1/2G e (p, ω)G e∗ (p − ζ, ω) exp[iωζX 2 ]. (8.1.4)
−p 0

In this form for the cross-correlation between the two stations from a
single source we are able to identify a phase component relating directly to
propagation between the stations exp[iωp X 12 ], which is modulated by a further
slowness integral.
With a broad distribution of many sources, the cross-correlation is summed
and the coherent part corresponding to the vicinity of the direct propagation
path is enhanced by constructive interference. The application of a stationary
phase treatment to the second integral over differential slowness in (8.1.4), as
in Snieder (2004), extracts the neighbourhood of ζ = 0, for which the slowness
of the arrivals that contribute to the net cross-correlation is the same at both
stations. In this case (8.1.4) reduces to a single integral over slowness. Full
suppression of extraneous contribution requires a good distribution of sources
relative to the inter-receiver path (e.g., Halliday & Curtis, 2008). However, with
a wavefield dominated by fundamental mode surface waves, well separated in
slowness from any other contributions, the requirements are less stringent.
The first integral over slowness in (8.1.4) includes a term exp[iωpδX 12 ] that
depends on δX 12 , the extent that the inter-receiver distance X 12 deviates from
the difference between the distances from each source to the two stations
|X 1 − X 2 |. With a sufficient distribution of sources this oscillatory term is
again suppressed by destructive interference leaving just contributions where
δX 12 ∼ 0, for which the paths from the effective sources to the two stations
are approximately aligned with the inter-receiver path. Two such zones are
present stretching out from the two receivers along the continuation of the
inter-receiver path.
Thus, provided the distribution of sources is sufficient to allow destructive
interference of incoherent phase effects, the summed cross-correlation over
many sources reduces to a form representing a virtual source-receiver pair at
8.2 Visualising cross-correlation effects 87

the two stations with contributions from propagation in each direction

­ 12 ® ω〈h 1 (ω)h 2 (ω)〉
Z p0
E (ω) = |M (ω)|2 dp p 1/2G e (p, ω)G e∗ (p, ω)
F (X ) 0
·{exp[iωp X 12 ] + exp[−iωp X 12 ]}. (8.1.5)

The expression (8.1.5) has a similar form to the time derivative of the
Green’s function for strain between the two stations, but the combination
G e (p, ω)G e∗ (p, ω) replaces G e (p, ω). The geometrical spreading term F (X ) will
not have a simple relation to the path, but will tend to be dominated by
source contributions from near the two stations, and hence F (X ) ∼ X 12 . The
averaged orientation factor 〈h 1 (ω)h 2 (ω)〉 will resemble the contribution for the
combination of the orientation at receivers 1 and 2 relative to the path between
them, because of the concentration around the inter-receiver path.
With low frequency approximations for the orientation effects for Rayleigh
and Love waves (3.3.6), the general expression (8.1.5) can be cast in a form
equivalent to that presented by Ji et al. (2024), also using a stationary phase
argument. They employ the factors appropriate to propagation just along the
path between the receivers, so that

A R = cos2 ψ1 cos2 ψ2 ; A L = 41 sin 2ψ1 sin 2ψ2 , (8.1.6)

where ψ1 , ψ2 are the angles between the inter-receiver path and the DAS cable
at each receiver.
Surface wave contributions come from the poles of the integrand in (8.1.5).
Their position in slowness, which controls dispersion, is unchanged from the
Green’s function but the pole is now second order (rather than first order for the
Green’s function) and so the amplitude factor is modified (Kennett & Fichtner,
2020). Typically the dominant contribution comes from fundamental mode
surface waves. Provided there are sources with a broad range of azimuths to the
inter-receiver path to contribute to the net cross-correlation, the contribution
corresponding to inter-receiver propagation is emphasised and the dispersion
for fundamental modes can readily be extracted.

8.2 Visualising cross-correlation effects

For a single frequency, we can develop a simple visualisation of the way in which
the cross-correlation field emerges from the contributions of many sources
(Figure 8.1). We show two receivers separated by 400 m in a simulation of
local conditions, look at the interaction of phase interference and geometrical
spreading from a grid of sources around the two receiver locations. We first
consider the scenario without orientation effects.
In Figure 8.1(a) we show the effect of wavefront spreading from the sources,
using an improved approximation to the spreading function that avoids the
singularities of the asymptotic form in (8.1.1). Because of the relatively fast drop
off with distance, sources close to the two receivers dominate. The constructive
88 Ambient noise correlation

Figure 8.1 Correlation simulation for 4 Hz waves with phase speed 500 m/s at stations
400 m apart, without allowance for directional effects. The positions of the stations are
shown by purple dots in each panel. (a) Geometric spreading effects from a source to
the two stations. (b) The ratio of the difference between the distance from each source
to the two stations and the inter-station path length. (c) Phase contributions to the
cross-correlation. (d) Total effect for two seismometers of the terms in (a), (b) and (c)
– amplified by 8. (e) Relative phase between the contributions from propagation from
a source to the two stations and the inter-station path. (f) Net effect of modulating the
phase term by geometrical spreading – amplified by 10.

interference conditions leading to (8.1.4) emphasise those source locations for

which the difference between the distance from the source to the two stations is
close to the inter-receiver distance. This effect is represented in Figure 8.1(b) by
plotting the ratio of the difference in distance to the path length |X 1 − X 2 |/X 12 ,
from zero (black) to unity (white).
With the specification of the frequency ω and slowness p c we show the phase
8.3 DAS orientation effects 89

contributions to the correlation using the function cos[ωp c (|X 1 − X 2 | − X 12 )]

as in Figure 8.1(c). We here use a frequency of 4 Hz and phase speed of
500 m/s (slowness 0.002 s/m), typical of many near-surface conditions. The
zones approximately in line with the stations show slow variation in phase
allowing constructive inference. But, as the inclination to the path between
the receivers increases, variation in phase is rapid and increasingly so at even
higher frequencies. It is the superposition of these rapidly varying phases that
leads to destructive interference for sources away from the path line and thus
the concentration of the cross-correlation on the inter-receiver path.
When we combine effects of the phase and spreading from the various
sources to a pair of receivers, the total result is as in Figure 8.1(d). The
distance-match term has here been applied as a multiplier to the product of
the geometric spreading and the phase variation. The dominant contribution
to the cross-correlation comes from beyond the stations along the extension of
the path between them, but some contamination is possible from sources lying
between the stations.
The application of the distance-match mask from Figure 8.1(b) produces a
net result on the correlation that has a strong resemblance to the source kernels
derived by Sager et al., (2017) for such inter-station correlations. In a similar
way we can simulate the equivalent structural kernel. Figure 8.1(e) displays the
phase factor cos[ωp c (X 1 + X 2 − X 12 )] representing the difference between the
phase accumulated in passage from the source to the two receivers compared
with that for the inter-receiver path. We modulate the phase pattern by the
geometric spreading effect from Figure 8.1(a) and the resultant, displayed in
Figure 8.1(g) puts strong emphasis on the zone in the immediate vicinity of the
inter-receiver path (cf. Sager et al. 2017).

8.3 DAS orientation effects

The results in Figure 8.1 are directly applicable to the case of a vertical
component seismograph where there is no directional sensitivity, but when
each of the receivers represents a DAS channel we need to take into account the
orientation of the cable relative to the impinging waves, as well as the potential
influence of gauge length averaging.
For a Rayleigh wave, with slowness p arriving with an inclination ψ relative to
the cable, the gauge-length effect from (3.3.4) is

2ω ³ ´
〈²̇d (ω)〉r = u r cos ψ sin 21 ωg p cos ψ . (8.3.1)
g

For a typical gauge length of 10 m and a phase speed of 500 m/s, for frequencies
less than 10 Hz the sine function does not impose much distortion. Because
strain is a second-order tensor quantity, the effect of inclination depends on
90 Ambient noise correlation

Figure 8.2 DAS orientation effects with in-line receivers 400 m apart: (a) Product of
Rayleigh wave orientation factors at the two receivers; (b) Net effect for Rayleigh waves
for in-line DAS sensors; (c) Product of Love wave orientation factors at the two receivers;
(d) Net effect for Love waves for in-line DAS sensors. The orientation of a 10 m gauge
length around each receiver channel is marked.

2ψ and to a good approximation for Rayleigh waves along a cable with uniform
orientation

〈²̇d (ω)〉r ∼ ω2 u r p cos2 ψ. (8.3.2)

In consequence there is a strong dependence on the position of any source.

For Love waves, the full gauge length effect is

2ω ³ ´
〈²̇d (ω)〉t = u t sin ψ sin 12 ωg p cos ψ ; (8.3.3)
g

with the low frequency approximation

〈²̇d (ω)〉t ∼ ω2 u t p sin ψ cos ψ. (8.3.4)

The most common local ambient noise is Rayleigh waves from anthropogenic
activities such as traffic, and Love waves are less frequently encountered since
they have a less favourable orientation effect. By bringing in the orientation
terms at each of the receivers we can examine the DAS case directly for both
Rayleigh and Love waves.
8.3 DAS orientation effects 91

Figure 8.3 DAS orientation effects with receivers 400 m apart on parallel cable
segments: (a) Product of Rayleigh wave orientation factors at the two receivers; (b) Net
effect for Rayleigh waves for parallel DAS sensors; (c) Product of Love wave orientation
factors at the two receivers; (d) Net effect for Love waves for parallel DAS sensors. The
orientation of a 10 m gauge length around each receiver channel is marked.

(a) In-line case

In Figure 8.2 we show the situation where both DAS channels lie on a cable
segment that aligns with the path between them.
The effect of the orientation factor cos2 ψ, for Rayleigh waves, is displayed
in Figure 8.2(a) and has a strong suppression effect for sources broadside to
either of the stations. This factor modulates the response for the case without
orientation effects shown in Figure 8.1(d) to give a total Rayleigh wave DAS
contribution shown in Figure 8.2(b), where there is a strong emphasis on
sources nearly in-line with the cable, beyond the receivers. The rapid phase
variations for sources between the receivers will lead to destructive interference
of Rayleigh wave contributions.
In contrast, for Love waves the orientation effects are completely
unfavourable for in-line propagation (Figure 8.2c) and very little coherent
energy is available in those part of the domain where the directional effects
would be favourable. The net result is almost no suitable source locations to
yield Love waves by correlation (Figure 8.2d).
(b) Parallel cable segments
We now consider the situation where two DAS channels are correlated that lie
92 Ambient noise correlation

Figure 8.4 DAS orientation effects with receivers 400 m apart on orthogonal cable
segments: (a) Product of Rayleigh wave orientation factors at the two receivers; (b)
Net effect for Rayleigh waves for orthogonal DAS sensors; (c) Product of Love wave
orientation factors at the two receivers; (d) Net effect for Love waves for orthogonal
DAS sensors. The orientation of a 10 m gauge length around each receiver channel is
marked.

the same 400 m distance apart but on parallel cable segments.

Now the interaction of the two orientation factors for Rayleigh waves gives
most emphasis to part of the zone between the cable segments (Figure 8.3a),
but this area shows rapid phase variations ((Figure 8.3b) and hence will be
unfavourable for extracting Rayleigh wave from correlation. A modest zone
near each receiver channel will have suitable coherence.
For Love waves, the product of the orientation factors is much more
favourable than in the in-line scenario (Figure 8.3c), and now there is a coherent
area along the line between the receiver segments (Figure 8.3d). This means that
this configuration with parallel cable segments will be suitable for extracting
Love waves if they are being excited by local sources.
(c) Orthogonal cable segments
Now we look at the situation where two DAS channels are correlated that lie the
same 400 m distance apart but on orthogonal cable segments.
The product of the orientation factors for Rayleigh waves (Figure 8.4a)
emphasises the zones beyond the receivers, with strong modulation of the
phase bands Figure 8.4b). Effective extraction of Rayleigh waves may be
8.4 Rayleigh and Love waves from DAS correlation 93

Figure 8.5 A record section of DAS cross-correlations with channel separation up to

250 m along a street in Bern, Switzerland. The orange lines indicate the acausal and
causal propagation of Rayleigh waves. The pink dashed line indicates the limit used
for interpretation of dispersion (100 m). The correlation signals become more complex
and difficult to reliably interpret beyond this distance. [Courtesy of Krystyna Smolinski]

possible, but different source areas will have opposing signs and so any net
effect will be reduced.
For Love waves the orientation product is similar to that for the parallel
segment case, but with opposite sign. The net result is that once again Love
waves can be extracted (Figure 8.4), but the polarity will be opposite to the
parallel segment situation.

8.4 Rayleigh and Love waves from DAS correlation

The cross-correlation of DAS channels on dark fibre has been successfully
employed by many different groups with the aim of extracting surface wave
dispersion, usually for Rayleigh waves with an in-line geometry. Even in such
a scenario, the 3-D complexity of the environment can limit the distance span
over which suitable dispersion measurements can be made.
For example, on the DAS experiment along a Bern street, for which noise
examples were illustrated in Chapter 5, the complexity of the correlation
wavefield limits analysis of Rayleigh wave dispersion to channel separations of
100 m or less for many parts of the line (Kennett et al., 2023). Along the DAS
cable, there are complex subsurface conditions with unconsolidated sediment
94 Ambient noise correlation

Figure 8.6 Illustration of the relative extraction of Rayleigh (R) and Love (L)
contributions from DAS channel cross-correlation depending on the relative
orientation of the cable at the receiver channels being correlated. In-line correlation
(left) is completely dominated by Rayleigh waves. Love waves start to emerge when the
channels have orthogonal direction (third panel). Love waves are most pronounced
when correlating parallel channels. The asymmetry in the correlation results is
associated with stronger noise from the area to the south and southwest of the cable
location including downtown San Jose. [After Ji et al., 2024].

on top of bedrock compounded by significant concrete infrastructure on top

lining the roadway. An example of correlations carried to 250 m channel
separation is shown in Figure 8.5, which indicates the presence of notable
secondary sources adding in their contributions to the primary Rayleigh wave
signal. These features are almost certainly related to the variation in the
structures below ground lining the road along which the DAS cables lies. There
are interruptions in the concrete cellar walls when side roads join, and also
other infrastructure embedded beneath the roadway. Careful choice of data, as
in Figure 8.5, has allowed successful characterisation of the subsurface structure
by inversion of Rayleigh wave dispersion (Smolinski et al., 2024).
A demonstration of the separation of Love and Rayleigh wave dispersion
has been made by Ji et al. (2024) from dark fibre in San Jose, California that
has three nearly linear cable legs, with two nearly parallel segments connected
by a nearly orthogonal link. In this experiment DAS correlation were made
using 3 days of data, stacking correlations over 1 min intervals. The gauge
length employed was 10 m, with strong background noise in the 1–10 Hz band.
For these frequencies the simpler approximations to orientation effects (8.3.2),
(8.3.4) will be adequate.
8.5 Mapping local heterogeneity 95

Examples of the results contrasting different modes of channel correlation are

shown in Figure 8.6, and are entirely consistent with the discussion above using
the visualisation of the correlation field. For in-line correlation, Rayleigh waves
are strongly expressed, and there is little hint of the presence of Love waves.
However, when correlation is made between channels on orthogonal segments
of cable a distinctive Love wave signals emerges with lower group speed. The
Love waves are especially clear when correlation is made between channels on
parallel cable segments, with polarity opposite to that when the cable runs are
orthogonal.
The corner configuration of the DAS cable allowed the separation of both
Love and Rayleigh wave dispersion, and successful joint inversion for local
structure.

8.5 Mapping local heterogeneity

DAS cables with long segments of straight runs are very suitable for extracting
estimates of surface wave dispersion from cross-correlation of DAS channels
using successive portions of the cable, with a suitable distribution of ambient
noise sources. Several groups working in urban areas with dark fibre have found
it beneficial to use time windows from the night when noise is more uniformly
distributed than in the heavier traffic conditions during the day (e.g., Spica et
al., 2020a; Smolinski et al., 2024).
On land, attention has been directed to Rayleigh wave dispersion (Zeng et al.,
2017; Spica et al., 2020a; Yang et al, 2022; Mirzanejad et al., 2024, Smolinski et
al., 2024). For marine cables, the Scholte waves linked to the ocean/sediment
interface have also been exploited successfully (Spica et al., 2020b; Liu et al,
2022). Averaging over segments of cable can help improve the rendering of
dispersion spectra, but comes at the price of reduced lateral resolution. Even
where closely spaced channels are employed, the gauge length used in the DAS
recording will place a lower limit on potential resolution.
Smolinski et al. (2024) discuss the complications imposed when working with
dark fibre in an urban environment, including the suppression of scattered
signals by f − k filtering so that dispersion spectra are not contaminated.
Where conditions are favourable, longer segments of cable can be employed
to improve the rendering of lower frequency dispersion, and thus deeper
penetration in the resulting model (Mirzanejad et al, 2024). The presence
of man-made infrastructure in the near surface, such as concrete and road
foundations complicates the rendering of dispersion, with thin high wavespeed
layers that modify the higher frequency dispersion.
Provided the geometry of the cable is well mapped, effective DAS channel
correlation is not confined to straight segments. Modest curvature can readily
be accommodated since, as we have seen above, the DAS field will emphasise
Rayleigh waves. Two-dimensional tomographic mapping from DAS correlation
96 Ambient noise correlation

of Rayleigh waves has been demonstrated by Fang et al. (2022) using a

rectangular array in a location with a strong microseismic field.
A common assumption made in the inversion of ambient noise dispersion
for horizontally varying structure is that the local dispersion curves correspond
directly to the vertical structural profile beneath that point. This approximation
is only valid when the horizontal variability is low (Woodhouse, 1974), and
depends on the independent propagation of the surface wave modes. This
condition is satisfied when there is just a slow evolution of the shapes of the
eigenfunctions associated with the modes (Kennett, 1995). Variations in shear
wavespeed have a modest effect, and slow variations in the depth of major
interfaces can be tolerated but sharp changes cannot. A slow progression in
the dispersion spectra for different locations should be fine. However, a sharp
change in the mode structure is a warning sign for problems that can result in
a biased representation of a zone of rapid change. Apparent dispersion across
a vertical interface can look satisfactory but prove difficult (or impossible) to
fit with a purely stratified model. Such zones need to be excised to avoid
severe distortion of the results. Indicators of strong lateral variability include
significant asymmetry in the character of the causal and acausal correlations
functions around a specific DAS channel (Yang et al., 2022).
For structures with strong variations, direct numerical approaches such as
full waveform inversion working directly with DAS strain-rate data (Liu et al.,
2022) offer promise for improved rendering, though they are computationally
intensive.
9
DAS array response

As we have seen in Chapter 3 a DAS system has strong local directivity effects
and is insensitive to waves arriving broadside to the cable. The high number
of recording channels available with DAS suggest that it should be possible to
use stacking methods to enhance coherent signal at higher frequencies, with
suppression of incoherent noise. Here we examine the way in which DAS cable
configurations can be used for signal analysis using array processing, taking
account of the dependence of the recorded response on the orientation of the
cable with respect to the passage of the seismic wavefield.
For conventional seismometers the directivity effects are equal for all sensors
and a single array response, determined by the layout of the sensors, can be
used to characterise the resolution of an array. The unidirectional nature of DAS
cables with local directivity effects mean that the response of a cable array has
to be evaluated directly for each plane wave component, rather than having a
universal function. As a result, the process of stacking an array of DAS channels
has a somewhat different character than for an array of seismometers, where
rotation of components is available. Such effects have been noted by, e.g.,
Lindsey et al. (2019), van den Ende & Ampuero (2021) in empirical analysis of
DAS stacking.
Näsholm et al. (2022) and Kennett (2022) have provided equivalent analyses
of the impact of directivity in DAS stacking and we build here on their results.

9.1 DAS array stacks

Array processing is designed to enhance coherent signal crossing an array by
making suitable combinations of the recorded traces. Distant events can be
represented by a dominant plane wave with slowness s. The timing of the
individual records from the each site in the array can be adjusted compensate
for the expected phase differences for this plane wave. With a sum over the set
of shifted traces, arrivals with slowness close to s are enhanced since these will
be coherent across the set of traces.
Consider a plane wave with vector slowness s, and a set of sensors locations

97
98 DAS array response

{x j }, relative to a reference site. The linear array sum for this plane wave, with
appropriate time delays at the N DAS channels, takes the form

1 XN
E (s, ω) = 〈²̇〉(x j , ω)e−iω[x j .s] , (9.1.1)
N j =1

in terms of the vector slowness s and the sensor coordinates x j For a single
plane wave travelling with an azimuth φ, and an inclination to the vertical i ,
the horizontal slowness vector s = s(cos φ sin i , sin φ sin i ).
As we have seen in Chapter 3 we can express the ground motion at frequency
ω crossing a DAS array as a superposition of plane wave components. For a wave
with horizontal slowness vector p at a distance ∆ from a source, the averaged
strain rate aligned along the local cable direction d is

2ω ³ ωg ´
〈ė d (p, ω)〉 = u (d.n) sin d.p eiωp∆ , (9.1.2)
g 2

including full gauge length effects (3.3.5). The term p d = d.p represents the
slowness of the plane wave resolved along the cable, and so p d = p cos ψ as a
function of the inclination of the cable ψ relative to the radial direction from
the source. The projection d.n depends on the polarization of the relevant wave
type.
The presence of the frequency dependent term involving gauge length g
imposes restrictions on the range of frequencies for which simple stacking is
effective. such issues are discussed in some detail by Näsholm et al. (2022).
For a distant event at moderate frequencies, when the variation of the plane
wave along the cable is slow compared to the gauge length, we can simplify the
expression for the averaged strain, as in (3.3.5),

〈ė d ((p), ω)〉 = ω2 u (d.n)(d.p)eiωp∆ . (9.1.3)

This averaged strain rate is equivalent to the ground acceleration due to the
plane wave resolved along the cable modulated by the slowness along the cable.
It is convenient to extract the phase contribution exp{iωp∆0 }, corresponding to
propagation from the reference point. Then phase effects arise from the DAS
channel separations {x j } from this reference point.
In the space–time domain, we then have

Z Z∞
〈²̇〉(x 1 , x 2 , ω) = ω2
dp 1 dp 2 (d(x).n)(d(x).p)u(p 1 , p 2 , ω)eiω[p 1 x1 +p 2 x2 ] , (9.1.4)
−∞

where u(p, w) is the displacement as recorded by a seismometer, and d(x) is the

local unit vector tangent to the cable. With this form for 〈²̇〉(x j , ω) the array sum
9.1 DAS array stacks 99

(9.1.1) performed for slowness s becomes

N Z Z∞
2
E (s, ω) = ω dp 1 dp 2 u(p 1 , p 2 , ω)(d(x j ).n)(d(x j ).p)eiω[(p 1 −s1 )x j 1 +(p 2 −s2 )x j 2 ] ,
X
j =1 −∞
Z Z∞ N
2
=ω dp 1 dp 2 u(p 1 , p 2 , ω) (d(x j ).n)(d(x j ).p)eiω[(p−s).x j ] ,
X
(9.1.5)
−∞ j =1

where the sum involves both the polarisation and the orientation of the cable
for each plane wave component in the wavefield. The time shifts depend on
the differential slowness p − s, between the slowness components of the actual
wavefield p and the stacking target slowness s. Whereas, the orientation factors
for DAS stacking depend solely on p.
This complex result for a DAS array is in strong contrast with the situation
for an array of seismometers, where the wavefield is just modulated by the
array response function S shifted to the differential slowness ∆s = p − s. The
seismometer stack
Z Z∞
2
S(s, ω) = ω dp 1 dp 2 u(p 1 , p 2 , ω)S (p − s, ω); (9.1.6)
−∞

the same functional form is derived irrespective of the slowness s and the
pattern in slowness space is dictated by the geometry of the array. The array
response function
N
S (∆s, ω) = eiω[∆s.x j ]
X
(9.1.7)
j =1

is a scaled version of the Fourier transform of a set of delta-functions placed at

the array positions with respect to wavenumber. This array response function
S (∆s, ω) can be characterised by calculating the response for a vertically
incident wave for which s 1 = s 2 = 0.
The term (d(x j ).p) in (9.1.5) means that the direct array sum for the DAS
cable depends on the slowness p. The specific behaviour therefore depends on
the nature of the incident wavefront. There is almost no response for vertical
incidence when the wave motion is perpendicular to the cable, but sensitivity
grows as the angle of incidence to the vertical increases.
From (3.3.6) the orientation factor associated with the radial component of
motion to the path from the source (P and SV waves) is cos2 ψ j where ψ j
is the angle between the local slowness vector p = (p 1 , p 2 ) and the tangent
to the cable at each site x j . This factor will always be less than unity and
thus stacking will tend to be less effective than for the similar configuration
of seismometers even with amplification by the magnitude of the horizontal
slowness. For the transverse component of motion (SH waves) the orientation
factor is cos ψ j sin ψ j and so can take both positive and negative values. The net
100 DAS array response

result is that direct stacking for SH waves does not focus on the target slowness.
Such an effect can be seen for the S stack for the simulated teleseismic event in
Figure 7.8.
With a distribution of seismometers many different styles of array layout can
be contemplated, but for a DAS cable all channels employed have to be directly
connected. This restricts the range of configurations that can be deployed.
We here consider the case of a single cable. Future DAS interrogators may
prove capable of handling multiple cables simultaneously, and then a wider
range of designs will become feasible. Already, some experiments use two
separate interrogators that allows more complex geometrical configurations to
be exploited if a common time base is available.
Although the array representation (9.1.5) is more complicated than the
seismometer case, it is not difficult to implement as a processing sequence with
a scan across a grid of horizontal slowness vectors. Näsholm et al. (2022) provide
a detailed work flow for such a scheme.
As suggested by Kennett (2022) and Näsholm et al. (2022), a very suitable
configuration for creating an array from a DAS cable is an Archimedean spiral
with polar equation r = aθ, which closely resembles a set of concentric circles
with gentle curvature using a single continuous fibre (Figure 9.1a). Such a spiral
has been shown to have good characteristics with even a limited number of
available sensors (Wathelet 2005, Kennett et al. 2015). Here, for simplicity, a
6 km long cable (a = 80 m) is considered with DAS channels extracted every 50
m, to give a total of 120 effective sensors along the spiral. From (9.1.5), the stack
response of the DAS array depends on the specifics of the incoming wavefront.
In Figure 9.1 the wavefront comes from a far regional event at 4 Hz with a
horizontal slowness of 10 s/◦ , i.e., an apparent velocity of 11.2 km/s. For a
surface P wavespeed of 4 km/s this would correspond to an incidence angle of
21◦ for P and an incident angle of 12◦ for S. The distribution of array power is
constructed as a function of slowness for both a set of seismometers with the
same spatial configuration (Figure 9.1b) and the DAS array for the radial and
tangential components from the source (Figure 9.1c,d).
Even with the modest size array of seismometers, the number of sensors
means that it possible to achieve strong targeting without any orientation
effects. Side bands are weak and well separated from the main response.
Once the DAS orientation effects are included (Figure 9.1c) there is still a good
concentration of the array response around the correct azimuth, but the peak is
displaced to larger slowness (as would be appropriate to a closer event). Hence
the DAS cable has a useful array response for P and SV waves that register on the
radial component from the source, provided the event is not too far away. For
the transverse component (SH waves) the orientation terms kill off the response
at the target slowness and weak lobes lie to the sides of radial peak (Figure 9.1d).
In Figure 9.2 we show the equivalent response for a variety of geometrical
9.1 DAS array stacks 101

2.0
(a) 0.50
(b)

1.5
St

1.0 0.25

0.5

0.0 0.00

-0.5
0

-1.0 -0.25

-1.5 -5

-2.0 -0.50
-2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 -0.50 -0.25 0.00 0.25 0.50 -10

Power [dB]
Geometry [km] Slowness [s/km]

0.50
(c) 0.50
(d)
-15
DAS-R DAS-T
-20
0.25 0.25

-25

0.00 0.00

-0.25 -0.25

-0.50 -0.50
-0.50 -0.25 0.00 0.25 0.50 -0.50 -0.25 0.00 0.25 0.50

Slowness [s/km] Slowness [s/km]

Figure 9.1 Simulation of array response for an Archimedean spiral including DAS
cable orientation effects at 4 Hz. (a) array geometry for 6 km cable with an incident
wavefront crossing the array with an azimuth of 300◦ and slowness 10.0 s/◦ for an event
at epicentral distance 20.3◦ ; (b) array response with no allowance for orientation (St),
equivalent to seismometer stack; (c) array response including orientation effects for
radial motion (DAS-R: for P-SV waves); (d) array response including orientation effects
for transverse motion (DAS-T: for SH waves). The response is shown as relative power
to the peak of the array response in (b), with the colour scale shown at the right. The
location of the true slowness vector is indicated with a open circle in (b)–(d).

figures, using again a 6 km cable and around 120 channels. We contrast a

triangle with a pentagon and a heptagon for a regional S wave with slowness
16.0 s/◦ and an azimuth of 160◦ . With the larger number of sides the side
bands become more ring like in the array response without directional effects
(St). Once the directivity effects are included for the radial component (DAS-R)
the main peak is weakened and the side bands are correspondingly lower. As
expected the location of the peak response is moved to larger slowness in a
similar manner for all three configurations. For the transverse component
102 DAS array response

2.0 0.50 0.50 0.50

(a) St DAS-R DAS-T
1.5

1.0 0.25 0.25 0.25

0.5

0.0 0.00 0.00 0.00

-0.5

-1.0 -0.25 -0.25 -0.25

-1.5

-2.0 -0.50 -0.50 -0.50

-2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 -0.50 -0.25 0.00 0.25 0.50 -0.50 -0.25 0.00 0.25 0.50 -0.50 -0.25 0.00 0.25 0.50
Geometry [km] Slowness [s/km] Slowness [s/km] Slowness [s/km]
2.0 0.50 0.50 0.50 0
(b) St DAS-R DAS-T
1.5
-5
1.0 0.25 0.25 0.25

0.5
-10

Power [dB]
0.0 0.00 0.00 0.00

-0.5 -15

-1.0 -0.25 -0.25 -0.25

-20
-1.5

-2.0 -0.50 -0.50 -0.50 -25

-2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 -0.50 -0.25 0.00 0.25 0.50 -0.50 -0.25 0.00 0.25 0.50 -0.50 -0.25 0.00 0.25 0.50

Geometry [km] Slowness [s/km] Slowness [s/km] Slowness [s/km]

2.0 0.50 0.50 0.50
(c) St DAS-R DAS-T
1.5

1.0 0.25 0.25 0.25

0.5

0.0 0.00 0.00 0.0

-0.5

-1.0 -0.25 -0.25 -0.25

-1.5

-2.0 -0.50 -0.50 -0.50

-2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 -0.50 -0.25 0.00 0.25 0.50 -0.50 -0.25 0.00 0.25 0.50 -0.50 -0.25 0.0 0.25 0.50

Geometry [km] Slowness [s/km] Slowness [s/km] Slowness [s/km]

Figure 9.2 Simulation of array responses including DAS cable orientation effects at
4 Hz. Three configurations are shown with a 6 km cable and an incident wavefront
crossing the array with an azimuth of 160◦ and slowness 16.0 s/◦ appropriate for
regional S waves. As in Figure 9.1 the array response is shown with out orientation
effects (St) and for DAS recording of the radial and transverse components (DAS-R,
DAS-T). The response is shown as relative power to the peak of the St response, with the
colour scale shown at the right. (a) triangular array; (b) pentagonal array; (c) heptagonal
array.

(DAS-T) only side band effects survive the influence of directivity and there is
no contribution near the true slowness.
For surface waves, which already have a large slowness, the distortion
introduced by the DAS array stacking is much less, and so stacking for Rayleigh
waves on the radial component can be expected to work well (Figure 9.3c:
DAS-R), but Love waves on the tangential component will be suppressed (Figure
9.3d: DAS-T). The simple triangular array has stronger sidelobes than the spiral
in Figure 9.1, as also seen in Figure 9.2.
Luo et al. (2020) have demonstrated that Rayleigh and Love wave separation
can be achieved by using orthogonal arms of DAS cable with addition and
subtraction of the contributions from the arms. Here we show how such ideas
can be generalised to other DAS array configurations.
In terms of the angle ψ at which a plane wave component crosses the cable,
the orientation factor for radial propagation (P, SV, Rayleigh) is cos2 ψ (3.3.9),
9.1 DAS array stacks 103

(a) (c) (d)

2.0 0.50 0.50

1.5
DAS-R DAS-T

1.0 0.25 0.25

0.5

0.0 0.00 0.00

-0.5
0

-1.0 -0.25 -0.25

-1.5 -5

-2.0 -0.50 -0.50

-2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 -0.50 -0.25 0.00 0.25 0.50 -0.50 -0.25 0.00 0.25 0.50 -10

Power [dB]
Geometry [km] Slowness [s/km] Slowness [s/km]

(b) (e) (f) -15

0.50 0.50 0.50

S DAS-RL DAS-TL
-20

0.25 0.25 0.25

-25

0.00 0.00 0.00

-0.25 -0.25 -0.25

-0.50 -0.50 -0.50

-0.50 -0.25 0.00 0.25 0.50 -0.50 -0.25 0.00 0.25 0.50 -0.50 -0.25 0.00 0.25 0.50

Slowness [s/km] Slowness [s/km] Slowness [s/km]

Figure 9.3 Simulation of array response including DAS cable orientation effects at 4 Hz,
for a triangular array with enhanced stacking of SH component: (a) array geometry for
6 km cable with an incident wavefront crossing the array with an azimuth of 285◦ and
slowness 30.0 s/◦ ; (b) array response with no allowance for orientation (St), equivalent
to seismometer stack; (c) array response including orientation effects for radial motion
(DAS-R: for P-SV waves); (d) array response including orientation effects for transverse
motion (DAS-T: for SH waves); (e) modified array response to suppress radial motion
(DAS-RL: for P-SV waves); (f) modified array response to enhance transverse motion
(DAS-TL: for SH waves). The response is shown as relative power to the peak of the
array response in (b), with the colour scale shown at the right. The location of the true
slowness vector is indicated with a open circle in (b)–(f ).

and for tangential propagation (SH, Love) is sin ψ cos ψ (3.3.10). If we multiply
each response by tan ψ we convert the radial term to sin ψ cos ψ, i.e., that seen
previously for the tangential, and convert the tangential term to sin2 ψ which
no longer has the sign change that leads to stack cancellation. Since tan ψ has a
singularity when cos ψ = 0 at ψ = 90◦ , it is necessary to regularize such tangent
multiplication in applying a modified stack. This can be done with exclusion of
the immediate neighbourhood of the singularity where a zero factor is applied.
The resulting modified stacks are shown in Figure 9.3(e),(f) for the same
triangular DAS configuration as before. The modified radial stack (Figure
9.3e: DAS-RL) is visually identical to the prior tangential stack in Figure 9.3(d),
though the remapping has some very slight effects from the regularization. The
modified tangential stack (Figure 9.3e: DAS-TL) achieves the goal of stacking
up Love waves at the desired slowness, but the side bands are magnified by
the slowness mapping in the DAS response. Hence, if there is good control
104 DAS array response

2.0
(a) 0.50
(b)

1.5 St
1.0 0.25

0.5

0.0 0.00

-0.5 0

-1.0 -0.25
-5
-1.5

-2.0 -0.50 -10

Power [dB]
-2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 -0.50 -0.25 0.00 0.25 0.50
Geometry [km] Slowness [s/km]
0.50
(c) 0.50
(d) -15

DAS-R* DAS-TL -20

0.25 0.25

-25

0.00 0.00

-0.25 -0.25

-0.50 -0.50
-0.50 -0.25 0.00 0.25 0.50 -0.50 -0.25 0.00 0.25 0.50
Slowness [s/km] Slowness [s/km]

Figure 9.4 Simulation of array response including DAS cable orientation effects at
4 Hz, for a hexagonal array with slowness correction and enhanced stacking of SH
component: (a) array geometry for 6 km cable with an incident wavefront crossing the
array with an azimuth of 60◦ and slowness 10.0 s/◦ ; (b) array response with no allowance
for orientation (St), equivalent to seismometer stack; (c) modified array response to
remove slowness bias from radial motion (DAS-R*): for P-SV waves); (d) modified array
response to enhance transverse motion (DAS-TL: for SH waves). The response is shown
as relative power to the peak of the array response in (b), with the colour scale shown
at the right. The location of the true slowness vector is indicated with a open circle in
(b)–(d).

on the characteristics of arriving surface waves in terms of azimuth and

approximate slowness, it is possible to extract Rayleigh waves by direct stacking
of a DAS array with suitable geometry (DAS-R), and Love waves with an azimuth
weighted stack using a regularized tangent weight (DAS-TL). For linear cable
runs, stacking effects will depend on their orientation relative to the arriving
wavefronts.
The success of the azimuthal weighting to enhance SH waves shown in Figure
9.3 suggests that the application of an inverse slowness weight in stacking DAS
9.2 DAS arrays for correlation analysis 105

records could help to improve the focusing of the array beam onto the correct
slowness for body waves, particularly P with relatively small slowness. To avoid
the singularity at vertical incidence p = 0, a regularization such as a water
level has to be applied to the inverse slowness weighting. In Figure 9.4 we
illustrate this approach for a hexagonal array, again using 6 km of cable. Here
we have applied a regularised inverse slowness weight for the radial component
(DAS-R*) and the azimuthal weighting for the transverse component to enhance
this stack response (DAS-TL). Regularisation with a water level set at slowness
0.1 s/km was applied for the inverse slowness weighting. The resultant DAS-R*
stack is well focussed on the true target, but still rather weak.
Local variations in cable coupling in DAS layouts and the influence of
small-scale heterogeneity (Singh et al., 2020) can have a significant effect on
local strain and hence the DAS signal. Van den Ende & Ampuero (2021) have
improved coherence in stacking of DAS signals from the Brady Geothermal site
by conversion to ground velocity, which requires local stacking and knowledge
of the slowness along the cable (itself a potential unknown).
Steerable arrays require sites where good signal coherency is achieved to
enhance signal strength. With the orientation effects of DAS recording large
numbers of abrupt changes of cable direction are undesirable since they
complicate the array response. Thus, not all DAS layouts can be expected to be
suitable for array analysis, and a larger size than many current configurations
may be needed to get well focused beams.

9.2 DAS arrays for correlation analysis

We have seen in Chapter 8 how the cross-correlation of channels along a DAS
cable is affected by the relative orientation of the cable segments relative to the
path between them. With specific DAS deployments, or situations where noise
sources are well characterised – such as traffic noise – it may be possible to adapt
experimental layouts to make good use of the sources and their directionality.
However, in the case of dark fibre the geometry of the cable configuration is
already fixed and so not under user control.
Many DAS cables are laid under roads or just beside them, as in the Bern
Street example in Chapter 8, and then traffic noise can be exploited directly with
signals aligning with the axis of the cable. However, a cable laid parallel to the
road but at some distance from the road itself largely picks up broadside signals
and the axial component is weak (Dou et al., 2017)). In these circumstances,
a perpendicular DAS cable could be used with the effective source being the
passage of vehicles through a rather narrow part of the road, thanks to the DAS
inclination factors (cf. Figure 8.2). Dou et al. (2017) have demonstrated that
such a perpendicular cable can be used for time lapse analysis.
Often the layout of DAS cables, as in dark fibre, is determined by the most
convenient geometries for telecommunication purposes and so the orientation
106 DAS array response

may not be ideal for seismic applications. It may be possible to compensate

to some extent with directional corrections, but it is probably preferable to
choose portions of the DAS cable for analysis that have the best orientation
(e.g., Fang et al., 2022). Liu et al (2022) have demonstrated using DAS channel
correlation across a rectangular cable layout to extract Rayleigh wave dispersion
and undertake 2-D tomographic mapping of the subsurface.
When the source of noise is not known or there are many different forms of
noise, such as traffic noise from many directions, there are a different set of
challenges. Zhao et al. (2023) have shown the use of nearly perpendicular cable
segments to separate both Rayleigh and Love wave signals by beamforming,
including the use of polarity corrections, with the object of characterising the
noise distribution. Here we aim to develop a DAS cable configuration that has
good overall behaviour for resolving the directional behaviour of the surface
waves that are the most common form of noise. For seismometer arrays,
the array response function S (9.1.7) provides a full summary of behaviour.
Whereas for DAS systems, as we have seen the strain-rate response is modulated
by the slowness p and the directional factors at each segment of the cable
depend on the specific incoming wave (9.1.5). The result is that the actual
response is distorted from the ideal expressed by S with bias toward larger
slowness. Even so the array response function can be a useful comparator for
different layouts (e.g., Kennett et al., 2023).
For each array configuration specified by the set of points {xj }, there is an
associated “co-array” Haubrich (1968) comprising the vectors
Xi j = xi − x j , i , j = 1, 2, . . . , N . (9.2.1)
In the context of inter-station correlation, the pattern of the co-array specifies
the sampling achievable. For optimum performance when using correlations,
we desire both a reasonable array response function and thorough sampling by
the co-array vectors.
For the design of small arrays, Maranò et al. (2014) have presented an
optimisation scheme with an objective function based on the character of the
array response. For the co-array, it is not obvious what would be the most
suitable criterion for any optimisation. Haubrich (1968) used a space-filling
approach based on direct search for a small number of sensors, but this is not
easy to generalise. Here we use the visual properties of the co-array as a guide
to behaviour.
When we turn to the design of DAS configurations with a broad range of noise
sources, we are faced with the topological necessity that all sensor locations
can be connected by a single cable. We again consider 6 km cables and select
36 elements, a sufficiently large number to show complex character but small
enough that the nature of the response can be readily appreciated.
The regular co-array is very helpful for assessing the potential of an array
of seismometers for cross-correlation analysis, but when we consider an
9.2 DAS arrays for correlation analysis 107

2.0 2.4 1.00

(a) R
1.5 1.8

1.0 1.2 0.75

0.5 0.6

0.0 0.0 0.50 0.50

L
-0.5 -0.6

-1.0 -1.2 0.25 0.25

-1.5 -1.8

-2.0 -2.4 0.00 0.00

-2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 -2.4 -1.8 -1.2 -0.6 0.0 0.6 1.2 1.8 2.4
Geometry [km] Inter-station vectors [km]
2.0 2.4 1.00
(b) R
1.5 1.8

1.0 1.2 0.75

0.5 0.6

0.0 0.0 0.50 0.50

L
-0.5 -0.6

-1.0 -1.2 0.25 0.25

-1.5 -1.8

-2.0 -2.4 0.00 0.00

-2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 -2.4 -1.8 -1.2 -0.6 0.0 0.6 1.2 1.8 2.4
Geometry [km] Inter-station vectors [km]
2.0 2.4 1.00
(c) R
1.5 1.8

1.0 1.2 0.75

0.5 0.6

0.0 0.0 0.50 0.50

L
-0.5 -0.6

-1.0 -1.2 0.25 0.25

-1.5 -1.8

-2.0 -2.4 0.00 0.00

-2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 -2.4 -1.8 -1.2 -0.6 0.0 0.6 1.2 1.8 2.4
Geometry [km] Inter-station vectors [km]

Figure 9.5 DAS array responses showing the geometrical layout with the cable
orientation at each sample point marked, the co-array behaviour with orientation
factors and the array function in slowness space for a 4 Hz signal. The cable length is 6
km in each case: (a) Archimedean spiral with 36 elements; (b) 36-element fan using a
continuous cable; (c) hexagonal layout with a single cable. The amplitudes of the R and
L orientation factors for Rayleigh and Love waves from (9.2.2) and (9.2.3) are plotted for
each vector in the co-array, with the L factor superimposed on the centre of the larger R
symbol. Stronger colours indicate the expectation of good cross-correlation results for
the wave type.
108 DAS array response

application to DAS arrays we also need to bear in mind the influence of cable
orientation relative to the path between the stations (Martin et al., 2021). Hence
we extract measures of merit for both Rayleigh and Love waves
For two DAS channels being correlated with angles ψ1 , ψ2 between the local
cable configuration and the path between the channel locations, the scaling
factor for Rayleigh waves due to the relative orientation is

R = cos2 ψ1 cos2 ψ2 , 0 ≤ R ≤ 1; (9.2.2)

and the equivalent factor for Love waves is

L = sin ψ1 cos ψ2 sin ψ2 cos ψ2 , − 12 ≤ L ≤ 12 . (9.2.3)

For each vector in the co-array we can associate these orientation factors and so
assess the effectiveness of the array design for the analysis of ambient Rayleigh
waves and incidentally examine where there may be the possibility of picking
up sensitivity to Love waves.
In the DAS co-array plots we display both the R and L factors for each
inter-channel vector. The R factors are shown in red and the inner, half-size,
L factors in cyan. The intensity of the colour indicates the level of potential
recovery, very weak factors are faded towards white. Thus distinctly visible
co-array vectors indicate the combinations of inter-station separation and
orientations for which good correlation results can be expected.
In Figure 9.5 we show three DAS configurations with reasonable properties
for both slowness response and co-array properties. Figure 9.5(a) shows the
use of 36 elements from an Archimedean spiral. Even where the cable does
not complete a full loop, the geometry of the co-array including the orientation
factors gives good azimuthal coverage for Rayleigh waves – though there is some
clumping in distance.
The second design shown in Figure 9.5(b) is a fan array with 36-elements.
This design is aimed to exploit the roughly 60◦ span around the vector between
DAS channels that will make an effective contribution to the correlation. With
6 such cable segments, and the links for external coupling, surprisingly good
properties are achieved. As might be expected the strongest Rayleigh factors are
associated with direct propagation along the arms of the fan, but reasonable
sensitivity for Rayleigh waves is achieved across a wide range of directions.
For such an array configuration, some mild irregularity in layout could also be
beneficial by spreading the range of azimuths. A similar ‘umbrella’ design for a
DAS layout has been suggested by van der Ende & Ampuero (2021), but they do
not provide any analysis of its performance.
We have seen in Figure 8.6 how the correlation between DAS channels on
parallel cable segments can be effective in enhancing Love waves from ambient
noise. The third design shown in Figure 9.5(c) attempts to exploit this property
with a hexagonal configuration that includes 3 sets of parallel cable segments.
9.2 DAS arrays for correlation analysis 109

In this case the Love response is more marked than for the other configurations
and can often be more prominent than for Rayleigh waves.
Frequently, that the Rayleigh wave response for the arrays dominates that for
Love waves, even when the DAS cable has a significant curvature. For the spiral
and the fan design the L factors remain quite small for most station pairs (Figure
9.5), so that Love wave contamination of Rayleigh wave results will only become
an issue if the local ambient noise has much stronger Love wave content.
For the hexagonal DAS layout it may be possible use azimuthal weighting
in combination with cross-correlation to effectively extract and identify Love
waves.
With a real DAS cable it is possible to use a much larger number of recording
positions than illustrated in Figure 9.5, so that the discrete spots will spread into
diffuse patches. It is also possible to select the portions of a DAS cable to be
used for correlation, so that poorly oriented segments or linking loops can be
excluded. Many dark fibres include segments with orthogonal orientation and
parallel segments with offset. If the offset is sufficient then the correlation on
parallel segments employed by Ji et al. (2024) will enable Love wave extraction
if suitable noise sources exist. Correlation along straight cable segments will
frequently produce good Rayleigh wave results.
10
Exploiting the DAS wavefield directly

It is always tempting to transform unfamiliar seismic records such as DAS strain

rate to an equivalent ground velocity that familiar tools can be applied. I hope to
have shown that once the nature of the DAS system and its effect on the seismic
wavefield are appreciated, there is much that can be done to exploit the wealth
of information provided by dense sampling.
Only modest modifications need to be made to computational methods
for seismogram synthesis to provide a direct representation of DAS records
including the influence of cable orientation and gauge-length effects. The
analysis of these effects is helpful in understanding the way in which DAS
systems record the different aspects of the seismic wavefield on local, regional
and global scales.
At close ranges the influence of gauge length is strong and can be included
by differencing ground velocity resolved along the fibre at the ends of the
gauge interval. For more distant events, the gauge averaging of strain rate is
equivalent to extracting the acceleration resolved along the cable modulated
by the horizontal slowness along the cable. This approximation works well
for typical seismic frequencies. Once the configuration of the cable has been
determined, with calibration of DAS channels against physical position, it is
possible to directly simulate the DAS signal using numerical integration of the
frequency-slowness response. For teleseismic body waves an effective approach
is to employ a bundle of slowness around the geometrical arrival so that depth
phase effects are correctly represented.
Not only does such synthesis of DAS records allow an analysis of the
different phase contributions to the observed wavefield with allowance for cable
orientation, it also has a valuable role in allowing assessment of coupling effects
along a cable, which are otherwise hard to determine. The relative simplicity of
far-field results means that DAS recordings can be used in waveform inversion
for source characteristics, alongside seismometer records.
The slowness domain representation of DAS recordings also provides a
convenient way of examining the behaviour of different configurations of DAS
cables. In particular, a single DAS layout can be employed as a steerable array

110
Exploiting the DAS wavefield directly 111

for high frequency waves. However, there is a tendency for the estimated
slownesses for body waves to be slightly enlarged by the slowness component
of the DAS response. Such effects are less important for surface waves, and by
suitable weighting both Rayleigh and Love waves can extracted from DAS stacks.
In this work I have concentrated on the use of horizontal DAS fibre
configurations, and have demonstrated the way in which the orientation
of the waves relative to the fibre has a strong control on the response.
Much work has been done on DAS cables deployed in vertical boreholes
particularly in exploration and production contexts. Borehole recordings at
typical seismic frequencies recordings can have significant value for analysing
many different types of seismic wave propagation phenomena in a lower noise
environment (e.g., Glubokovskikh et al., 2021). The analysis used here in terms
of the frequency–slowness domain can be extended to downhole studies by
modification of modelling codes for VSP simulation with differencing in the
vertical direction, and orientation factors relative to the vertical rather than the
horizontal.
Bibliography

Ajo-Franklin JB, Dou S, Lindsey NJ, Monga I, Tracy C, Robertson M, Tribaldos VR,
Ulrich C, Freifeld B, Daley T, Li X (2019) Distributed acoustic sensing using dark
fibre for near-surface characterization and broadband seismic event detection.
Scientific Reports, 9(1), 1–14. doi: 10.1038/s41598-018-36675-8
Baird AF, Stork AL, Horne SA, Naldrett G, Kendall J-M, Wookey J, Verdon JP, Clarke A
(2020) Characteristics of microseismic data recorded by distributed acoustic
sensing systems in anisotropic media. Geophysics, 85, KS139–KS147,
doi: 10.1190/geo2019-0776.1
Bakku S (2015) Fracture characterization from seismic measurements in a borehole,
Ph.D. Thesis, Mass. Inst. Technol., Cambridge, MA.
Bakulin A, Silvestrov I, Pevzner R (2020) Surface seismics with DAS: An emerging
alternative to modern point-sensor acquisition. The Leading Edge, 39, 808–818.
doi: 10.1190/tle39110808.1
Bensen GD, Ritzwoller MH, Barmin MP, Levshin AL, Lin F, Moschetti MP, Shapiro NM,
Yang Y (2007) Processing seismic ambient noise data to obtain reliable
broad-band surface wave dispersion measurements. Geophys. J. Int., 169(3),
1239–1260. doi: 10.1111/j.1365-246X.2007.03374.x
Binder G, Titov A, Liu Y, Simmons J, Tura A, Byerley G, Monk D (2020) modeling the
seismic response of individual hydraulic fracturing stages observed in a
time-lapse distributed acoustic sensing vertical seismic profiling survey.
Geophysics, 85, T225–T235. doi: 10.1190/geo2019-0819.1
Biondi E, Wang X, Williams EF, Zhan Z (2023) Geolocalization of large-scale DAS
channels using a GPS-tracked moving vehicle. Seism. Res. Lett, 94, 318–330.
doi: 10.1785/0220220169
Bock G, Denham D (1983) Recent earthquake activity in the Snowy Mountains region
and its relationship to major faults, J. Geol. Soc. Australia, 30, 423-429. doi:
10.1080/00167618308729267
Chapman CH (1973) The Earth flattening approximation in body wave theory.
Geophys. J. R. Astr. Soc., 35, 55–70. doi: 10.1111/j.1365-246X.1973.tb02414.x
Chen J, Pan L, Li Z, Chen X (2022) Continental reworking in the eastern South China
Block and its adjacent areas revealed by F-J multimodal ambient noise
tomography. J. Geophys. Res. Solid Earth, 127, e2022JB024776.
doi: 10.1029/2022JB024776

112
Bibliography 113

Chen X (1993) A systematic and efficient method of computing normal modes for
multilayered half-space. Geophys. J. Int., 115(2), 391–409.
doi: 10.1111/j.1365-246X.1993.tb01194.x
Cheng F, Xia J, Zhang K, Zhou C, Ajo-Franklin JB (2021) Phase-weighted slant stacking
for surface wave dispersion measurement. Geophys. J. Int., 226(1), 256–269.
doi: 10.1093/gji/ggab101
Cheng F (2024) Photonic Seismology: A New Decade of Distributed Acoustic Sensing
in Geophysics from 2012 to 2023. Surv. Geophys., 45(4).
doi: 10.1007/s10712-024-09840-0
Chmiel M, Mordret A, Boué P, Brenguier F, Lecocq T, Courbis R, Hollis D, Campman X,
Romijn R, Van der Veen W (2019) Ambient noise multimode Rayleigh and Love
wave tomography to determine the shear velocity structure above the Groningen
gas field. Geophys. J. Int., 218(3), 1781–1795. doi: 10.1093/gji/ggz237
Daley TM, Miller DE, Dodds K, Cook P, Freifeld BM (2016) Field testing of modular
borehole monitoring with simultaneous distributed acoustic sensing and
geophone vertical seismic profiles at Citronelle, Alabama. Geophysical
Prospecting, 64, 1318–1334. doi: 10.1111/1365-2478.12324
Dal Moro G (2020) The magnifying effect of a thin shallow stiff layer on Love waves as
revealed by multi-component analysis of surface waves. Sci. Rep., 10, 9071.
doi: 10.1038/s41598-020-66070-1
Dean T, Cuny T, Hartog AH (2017) The effect of gauge length on axially incident
P-waves measured using fibre optic distributed vibration sensing. Geophysical
Prospecting, 65, 184–193, doi: 10.1111/1365-2478.12419
Dong S, Li Z, Chen X, Fu L (2021) DisperNet: An effective method of extracting and
classifying the dispersion curves in the frequency–Bessel dispersion spectrum.
Bull. Seismol. Soc. Am., 111, 3420–3431, 2021. doi: 10.1785/0120210033
Dou S, Lindsey N, Wagner A, Daley TM, Freifeld B, Robertson M, Peterson J, Ulrich C,
Martin ER, Ajo-Franklin JB (2017) Distributed acoustic sensing for seismic
monitoring of the near surface: A traffic-noise interferometry case study. Sci. Rep.,
7, 11620. doi: 10.1038/s41598-017-11986-4
Dougherty SL, Cochran ES, Harrington RM (2019) The LArge-n Seismic Survey in
Oklahoma (LASSO) Experiment. Seism. Res. Lett., 90(5), 2051–2057.
doi: 10.1785/0220190094
Egorov A, Correa J, Bóna A, Pevzner R, Tertyshnikov K, Glubokovskikh S, Puzyrev V,
Gurevich B (2018) Elastic full waveform inversion of vertical seismic profile data
acquired with distributed acoustic sensors. Geophysics, 83, R273–R281.
doi: 10.1190/geo2017-0718.1
Fang J, Yang Y, Shen Z, Biondi E, Wang X, Williams EF, Becker MW, Eslamian D, Zhan Z
(2022) Directional Sensitivity of DAS and Its Effect on Rayleigh-Wave Tomography:
A Case Study in Oxnard, California. Seism. Res. Lett., 94(2A), 887–897. doi:
10.1785/0220220235
Fenta MC, Potter DK, Szanyi J (2021) Fibre optic methods of prospecting: A
comprehensive and modern branch of geophysics. Surv. Geophys., 42, 551–584.
doi: 10.1007/s10712-021-09634-8
Fichtner A, Tsai V (2019) Theoretical Foundations of Noise Interferometry, In Seismic
Ambient Noise (pp109–143), Cambridge University Press.
doi: 10.1017/9781108264808.006
114 Bibliography

Fichtner A, Hofstede C, Kennett BLN, Nymand N, Lauritzen ML, Zigone D, Eisen O

(2023) Fiber-optic airplane seismology on the Northeast Greenland Ice Stream.
The Seismic Record, 3(2), 125–133. doi: 10.1785/0320230004
Finger C, Löer K (2024) Depth of sudden velocity changes derived from multi-mode
Rayleigh waves. J. Geophys. Res,: Solid Earth, 129, e2023JB028322.
doi: 10.1029/2023JB028322
Gabai H, Eyal A (2016) On the sensitivity of distributed acoustic sensing. Optics Lett.,
41(24), 5648–5651. doi: 10.1364/OL.41.005648
Glubokovskikh S, Pevzner R, Sidenko E, Tertyshnikov K, Gurevich B, Shatalin S,
Slunyaev A, Pelinovsky E (2021) Downhole distributed acoustic sensing provides
insights into the structure of short-period ocean-generated seismic wavefield. J.
Geophys. Res.: Solid Earth, 126, e2020JB021463. doi: 10.1029/2020JB021463
Gök R, Walter WR, Barno J, Downie C, Mellors RJ, Mayeda K, Roman-Nieves J,
Templeton D, Ajo-Franklin J (2024) Reliable earthquake source parameters using
distributed acoustic sensing data derived from coda envelopes. Seism. Res. Lett.,
95, 2208–2220 doi: 10.1785/0220230270.
Haddon RAW (1984) Computation of synthetic seismograms in layered earth models
using leaking modes. Bull. Seism. Soc. Am., 74(4), 1225–1248.
doi: 10.1785/bssa0740041225
Halliday D, Curtis A (2008) Seismic interferometry, surface waves and source
distribution. Geophys. J. Int., 175(3), 1067–1087.
doi: 10.1111/j.1365-246X.2008.03918.x
Hartog AH (2017) An Introduction to Distributed Optical Fibre Sensors, CRC Press, Boca
Raton, FL.
Haubrich RA (1968) Array design. Bull. Seism. Soc. Am., 58(3), 977–991.
doi: 10.1785/BSSA0580030977
Hudson TS, Baird AF, Kendall J-M, Kufner SK, Brisbourne AM, Smith AM, Butcher A,
Chalari A, Clarke A (2021) Distributed acoustic sensing (DAS) for natural
microseismicity studies: a case study from Antarctica. J. Geophys. Res.: Solid Earth,
126, e2020JB021493. doi: 10.1029/2020JB021493
Ji Q, Luo B, Biondi B (2024) Exploiting the potential of urban DAS grids: Ambient-noise
subsurface imaging using joint Rayleigh and Love Waves. Seismol. Res. Lett., 95,
1794–1811. doi: 10.1785/0220230104.
Jiang C, Denolle MA (2020) NoisePy: A new high-performance Python tool for
ambient-noise seismology. Seism. Res. Lett., 91(3), 1853–1866.
doi: 10.1785/0220190364
Jiang C, Denolle MA (2022) Pronounced seismic anisotropy in Kanto sedimentary
basin: A case study of using dense arrays, ambient noise seismology, and
multi-modal surface-wave imaging. J. Geophys. Res. Solid Earth, 127(8),
e2022JB024613. doi: 10.1029/2022JB024613
Jousset P, Reinsch T, Ryberg T, Blanck H, Clarke A, Aghayev R, Hersir GP, Henninges J,
Weber M, Krawczyk CM (2018) Dynamic strain determination using fibre-optic
cables allows imaging of seismological and structural features. Nature
Communications, 9(1), 2509. doi: 10.1038/s41467-018-04860-y
Bibliography 115

Katakami S, Noda S, Korenaga M, Araki E, Takahashi N, Iwata N (2024) Potential of

earthquake strong motion observation utilizing a linear estimation method for
phase cycle skipping in distributed acoustic sensing. J. Geophys. Res.: Solid Earth,
129, e2023JB027327. doi: 10.1029/2023JB027327
Karrenbach M, Cole S, Ridge A, Boone K, Kahn D, Rich J, Silver K, Langton D (2019 )
Fibre-optic distributed acoustic sensing of microseismicity, strain and
temperature during hydraulic fracturing. Geophysics, 84, D11–D23.
doi: 10.1190/geo2017-0396.1
Kennett BLN (1980) Seismic waves in a stratified half space II - Theoretical
seismograms. Geophys. J. R. astr. Soc., 61, 1–10.
doi: 10.1111/j.1365-246x.1980.tb04299.x
Kennett BLN (1983) Seismic Wave Propagation in Stratified Media, Cambridge
University Press.
Kennett BLN, Clarke TJ (1983) Rapid surface wave dispersion calculations. Geophys. J.
R. Astr. Soc., 72, 633–645. doi: 10.1111/j.1365-246x.1983.tb02823.x
Kennett BLN (1985) On regional S. Bull. Seism. Soc. Am, 75, 1077–1088.
doi: 10.1785/BSSA0750041077
Kennett BLN, Engdahl ER, Buland R (1995) Constraints on seismic velocities in the
Earth from travel times. Geophys. J. Int., 122, 108–124.
doi: 10.1111/j.1365-246x.1995.tb03540.x
Kennett BLN (1995) Approximations for surface wave propagation in laterally varying
media. Geophys. J. Int., 122, 470–478. doi: 10.1111/j.1365-246x.1995.tb07008.x
Kennett BLN, Marson-Pidgeon K, Sambridge M (2000) Seismic source characterization
using a Neighbourhood Algorithm. Geophys. Res. Lett., 27, 3401–3404. doi:
10.1029/2000gl011559
Kennett BLN (2001) The Seismic Wavefield I: Introduction and Theoretical
Development, Cambridge University Press, Cambridge.
Kennett BLN (2002) The Seismic Wavefield II: Interpretation of Seismograms on
Regional and Global Scales, Cambridge University Press, Cambridge.
Kennett BLN, Stipčević J, Gorbatov A (2015) Spiral arm seismic arrays. Bull. Seism. Soc.
Am., 105, 2109–2116. doi: 10.1785/0120140354
Kennett BLN, Fichtner A (2020) Exploiting Seismic Waveforms: Correlation,
Heterogeneity and Inversion, Cambridge University Press. doi:
10.1017/9781108903035
Kennett BLN (2022) The seismic wavefield as seen by distributed acoustic sensing
arrays: local, regional and teleseismic sources. Proc. R. Soc. A, 478, 20210812.
doi: 10.1098/rspa.2021.0812
Kennett BLN (2023a) Interacting seismic waveguides: Multimode surface waves and
leaking modes. Seismica, 2(1). doi: 10.26443/seismica.v2i1.282
Kennett BLN (2023b) Construction of frequency–slowness and frequency–phase speed
diagrams. doi: 10.5281/zenodo.7384357.
Kennett B, Jiang C, Smolinski K (2023) Local station correlation: large N-arrays and
DAS. Seismica, 2(2). doi: 10.26443/seismica.v2i2.389
Kennett BLN, Lai VH, Miller MS, Bowden D, Fichtner A (2024) Near-source effects on
DAS recording: implications for tap tests. Geophys. J. Inter., 237, 436–444.
doi: 10.1093/gji/ggae055
116 Bibliography

Kerry NJ, (1981) Synthesis of seismic surface waves. Geophys. J. R. Astr. Soc., 64,
425–466. doi: 10.1111/j.1365-246X.1981.tb02675.x
Kuvshinov B (2016) Interaction of helically wound fibre-optic cables with plane
seismic waves. Geophysical Prospecting, 64, 671–688.
doi: 10.1111/1365-2478.12303
Lancelle, CE, Baldwin, JA, Lord, N, Fratta, D, Chalari, A, Wang, HF (2021) Using
distributed acoustic sensing (DAS) for multichannel analysis of surface waves
(MASW), In Distributed Acoustic Sensing in Geophysics (pp. 213–228), American
Geophysical Union (AGU) doi: 10.1002/9781119521808.ch16
Li J, Zhu W, Biondi E, Zhan Z (2023a) Earthquake focal mechanisms with distributed
acoustic sensing. Nature Commun., 14, 4181. doi: 10.1038/s41467-023-39639-3
Li J, Kim T, Lapusta N, Biondi E, Zhan Z (2023b) The break of earthquake asperities
imaged by distributed acoustic sensing. Nature, 620, 800–806. doi:
10.1038/s41586-023-06227-w
Li Z, Zhan Z (2018) Pushing the limit of earthquake detection with distributed acoustic
sensing and template matching: A case study at the Brady geothermal field.
Geophys. J. Int., 215, 1583–1593. doi: 10.1093/gji/ggy359
Li Z, Shi C, Chen X, (2021) Constraints on crustal P wave structure with leaking mode
dispersion curves. Geophysical Research Letters, 48, e2020GL091782.
doi: 10.1029/2020GL091782
Li Z, Shi C, Ren H, Chen X (2022) Multiple leaking mode dispersion observations and
applications from ambient noise cross-correlation in Oklahoma. Geophysical
Research Letters, 49, e2021GL096032, 2022. doi: 10.1029/2021GL096032
Li Y, Karrenbach M, Ajo-Franklin JB (Eds.) (2022) Distributed Acoustic Sensing in
Geophysics: Methods and Applications, AGU Geophysical Monograph, 238, doi:
10.1002/9781119521808
Lindsey NJ, Martin ER, Dreger DS, Freifeld B, Cole S, James SR, Biondi BL, Ajo-Franklin
JB (2017) Fiber-optic network observations of earthquake wavefields, Geophys.
Res. Lett., 44, 11,711–11,799. doi: 10.1002/2017GL075722.
Lindsey NJ, Dawe TC, Ajo-Franklin JB (2019) Illuminating seafloor faults and ocean
dynamics with dark fiber distributed acoustic sensing. Science, 366(6469),
1103–1107. doi: 10.1126/science.aay5881
Lindsey NJ, Rademacher H, Ajo-Franklin JB (2020) On the broadband instrument
response of fiber-optic DAS arrays. J. Geophys. Res.: Solid Earth, 125,
e2019JB018145. doi: 10.1029/2019JB018145
Lindsey NJ, Martin ER (2021) Fiber-optic seismology. Annu. Rev. Earth Planet Sci.,
49(1), 309–336. doi: 10. 1146/annurev-earth-072420-065213
Lior I, Sladen A, Rivet D, Ampuero J-P, Hello Y, Becerril C, Martin HF, Lamare P, Jestin C,
Tsagkli S, Markou C (2021) On the detection capabilities of underwater distributed
acoustic sensing. J. Geophys. Res.: Solid Earth, 126, e2020JB020925. doi:
10.1029/2020JB020925
Liu H, Li J, Wang Y (2022) Wave equation dispersion inversion of distributed acoustic
sensing data. J. Geophys.l Res. Solid Earth, 127, e2022JB024671.
doi: 10.1029/2022JB024671
Lobkis OI, Weaver RL (2001) On the emergence of the Green’s function in the
correlations of a diffuse field. J. Acoust. Soc. Am., 110(6), 3011–3017.
doi: 10.1121/1.1417528
Bibliography 117

Luo B, Trainor-Guitton W, Bozdaǧ W, LaFlame L, Cole S, Karrenbach M (2020)

Horizontally orthogonal distributed acoustic sensing array for earthquake- and
ambient-noise-based multichannel analysis of surface waves. Geophys. J. Int., 222,
2147–2161. doi: 10.1093/gji/ggaa293
Maranò S, Fäh D, Lu YM (2014) Sensor placement for the analysis of seismic surface
waves: sources of error, design criterion and array design algorithms. Geophys. J.
Int., 197(3), 1566–1581. doi: 10.1093/gji/ggt489
Marson-Pidgeon K, Kennett BLN (2000) Flexible computation of teleseismic synthetics
for source and structural studies. Geophys. J. Int., 143, 689–699.
doi: 10.1046/j.1365-246x.2000.00268.x
Martin ER, Castillo CM, Cole S, Sawasdee PS, Yuan S, Clapp R, Karrenbach M, Biondi
BL (2017) Seismic monitoring leveraging existing telecom infrastructure at the
SDASA: Active, passive, and ambient-noise analysis. The Leading Edge, 36(12),
1025–1031. doi: 10.1190/tle36121025.1
Martin E (2018) Passive imaging and characterization of the subsurface with distributed
acoustic sensing. Ph.D. Thesis, Stanford Univ., Stanford, CA
Martin ER, Lindsey NJ, Ajo-Franklin JB, Biondi BL (2021) Introduction to
Interferometry of Fiber-Optic Strain Measurements, In Distributed Acoustic
Sensing in Geophysics (pp. 111–129).
American Geophysical Union (AGU) doi: 10.1002/9781119521808.ch9
Matsumoto H, Araki E, Kimura T, Fujie G, Shiraishi K, Tonegawa T, Obana K, Arai R,
Kaiho Y, Nakamura Y, Yokobiki T, Kodaira S, Takahashi N, Ellwood R, Yartsev V,
Karrenbach M (2021) Detection of hydroacoustic signals on a fiber-optic
submarine cable. Scientific Reports, 11(1), 2797. doi: 10.1038/s41598-021-82093-8.
Mestayer J, Cox B, Wills P, Kiyashchenko D, Lopez J, Costello M, Bourne S, Ugueto S,
Lupton R, Solano G (2011) Field trials of distributed acoustic sensing for
geophysical monitoring, SEG Technical Program Expanded Abstracts 2011, Society
of Exploration Geophysicists, 4253–4257.
Mirzanejad M, Seylabi E, Tyler S, Ajo-Franklin J, Hatch-Ibarra R, Saltiel S (2024)
Subsurface imaging using interferometry of distributed acoustic sensing ambient
noise measurement along a dark fiber line: A case study in downtown Reno,
Nevada. Bull. Seism. Soc. Am., 114, 1613–1625 doi: 10.1785/0120230136
Nakata N, Gualtieri L, Fichtner A (Eds.) (2019) Seismic Ambient Noise, Cambridge
University Press. doi: 10.1017/9781108264808
Näsholm SP, Iranpour K, Wuestefeld A, Dando BDE, Baird AF, Oye V (2022) Array signal
processing on daistribute acoustic sensing data: directivity effects in slowness
space. J. Geophys. Res. Solid Earth, 127(2), e2021JB023587.
doi: 10.1029/2021JB023587
Nayak A, Thurber CH (2020) Using multicomponent ambient seismic noise
cross-correlations to identify higher mode Rayleigh waves and improve dispersion
measurements,. Geophys. J. Int., 222(3), 1590–1605. doi: 10.1093/gji/ggaa270
Nayak A, Ajo-Franklin JB, the Imperial Valley Dark Fiber Team, (2021) Distributed
acoustic sensing using dark fiber for array detection of regional earthquakes.
Seismol. Res. Lett. 92, 2441–2452, doi: 10.1785/0220200416
118 Bibliography

Paitz P, Edme P, Gräff D, Walter F, Doetsch J, Chalari A, Schmelzbach C, Fichtner A

(2020) Empirical investigations of the instrument response for distributed
acoustic sensing (DAS) across 17 Octaves. Bull. Seismol. Soc. Am. 111, 1–10.
doi: 10.1785/0120200185
Park CB, Miller RD, Xia J (1999) Multichannel analysis of surface waves. Geophysics,
64(3), 800–808. doi: 10.1190/1.1444590
Parker T, Shatalin S, Farhadiroushan M (2014) Distributed Acoustic Sensing – a new
tool for seismic applications. First Break, 32, 61–69.
doi: 10.3997/1365-2397.2013034
Parker LM, Thurber CH, Zeng X, Li P, Lord NE, Fratta D, Wang HF, Robertson MC,
Thomas AM, Karplus MS, Nayak A, Feigl KL (2018) Active-source seismic
tomography at the Brady Geothermal Field, Nevada, with dense nodal and
fiber-optic seismic arrays,. Seism. Res. Lett. 89, 1629–1640. doi:
10.1785/0220180085
Quiros DA, Brown LD, Kim D (2016) Seismic interferometry of railroad induced ground
motions: body and surface wave imaging. Geophys. J. Int., 205(1), 301–313. doi:
10.1093/gji/ggw033
Rivet D, de Cacqueray B, Sladen A, Roques A, Calbris G (2021) Preliminary assessment
of ship detection and trajectory evaluation using distributed acoustic sensing on
an optical fiber telecom cable. J. Acoust. Soc. Am., 149(4), 2615–2627.
doi: 10.1121/10.0004129.
Roth M, Holliger K, Green AG (1998) Guided waves in near-surface seismic surveys.
Geophysical Research Letters, 25, 1071–1074. doi: 10.1029/98gl00549
Sager K, Ermert L, Boehm C, Fichtner A (2017) Towards full waveform ambient noise
inversion. Geophys. J. Int., 212(1), 566–590. doi: 10.1093/gji/ggx429
Salmon M, Kennett BLN, Saygin E (2013) Australian Seismological Reference Model
(AuSREM): crustal component. Geophys. J. Int., 192, 190–206.
doi: 10.1093/gji/ggs004
Shi C, Ren H, Li Z, Chen, X, (2021) Calculation of normal and leaky modes in planar
waveguides based on a semi-analytical spectral element method. Geophys. J. Int.,
230, 1928–1947. doi: 10.1093/gji/ggac163
Singh S, Capdeville Y, Igel H (2020) Correcting wavefield gradients for the effects of
local small-scale heterogeneities. Geophys. J. Int., 220, 996–1011.
doi: 10.1093/gji/ggz479
Sladen A, Rivet D, Ampuero J-P, De Barros L, Hello Y, Calbris G, Lamare P (2019)
Distributed sensing of earthquakes and ocean-solid Earth interactions on seafloor
telecom cables. Nature Communications, 10(1), 5777.
doi: 10.1038/s41467-019-13793-z
Snieder R (2004) Extracting the Green’s function from the correlation of coda waves: A
derivation based on stationary phase. Phys. Rev. E, 69(4), 046610.
doi: 10.1103/PhysRevE.69.046610
Song Z, Zeng X, Wang B, Yang J, Li X, Wang HF (2021) Distributed acoustic sensing
using a large-volume airgun source and internet fiber in an urban area. Seismol.
Res. Lett. 92, 1950–1960, doi: 10.1785/0220200274
Smolinski KT, Bowden DC, Paitz P, Kugler F, Fichtner A (2024) Shallow subsurface
imaging using challenging urban DAS data. Seism. Res. Lett., 95, doi:
10.1785/0220240052
Bibliography 119

Spica ZJ, Perton M, Martin ER, Beroza GC, Biondi B (2020) Urban seismic site
characterization by fiber-optic seismology. J. Geophys. Res. Solid Earth, 125, 1–14.
doi: 10.1029/2019JB018656
Spica ZJ, Nishida K, Akuhara T, Pétrélis F, Shinohara M, Yamada T (2020) Marine
sediment characterized by ocean-bottom fiber-optic seismology. Geophys. Res.
Lett., 47, e2020GL088360. doi: 10.1029/2020GL088360
Sun C, Wang Z, Wu D, Cai R, Wu H (2021) A unified description of surface waves and
guided waves with relative amplitude dispersion maps. Geophys. J. Int., 227,
1480–1495. doi: 10.1093/gji/ggab284
Udd E, Spillman WB (Eds.) (2024) Fiber optic sensors: an introduction for engineers and
scientists, 3rd edn., Wiley.
van den Ende M, Ampuero J-P (2021) Evaluating seismic beamforming capabilities of
distributed acoustic sensing arrays. Solid Earth 12, 915–934.
doi: 10.5194/se-12-915-2021
van Wijk K, Mikesell TD, Schulte-Pelkum V, Stachnik J (2011) Estimating the
Rayleigh-wave impulse response between seismic stations with the cross terms of
the Green tensor. Geophys. Res. Lett., 38(16), L16301. doi: 10.1029/2011GL047442
Walter F, Gräff D, Lindner F, Paitz P, Köpfli M, Chmiel M, Fichtner A (2020) Distributed
acoustic sensing of microseismic sources and wave propagation in glaciated
terrain. Nature Communications, 11, 2436. doi: 10.1038/s41467-020-15824-6
Wang HF, Zeng X, Miller DE, Fratta D, Feigl KL, Thurber CH, Mellors RJ (2018) Ground
motion response to an ML 4.3 earthquake using co-located distributed acoustic
sensing and seismometer arrays. Geophys. J. Int., 213(3), 2020–2036.
doi: 10.1093/gji/ggy102
Wapenaar K (2004) Retrieving the elastodynamic Green’s function of an arbitrary
inhomogeneous medium by cross correlation. Phys. Rev. Lett., 93(25), 254301.
doi: 10.1103/PhysRevLett.93.254301
Wapenaar K, Fokkema J (2006) Green’s function representations for seismic
interferometry. Geophysics, 71(4), SI33–SI46. doi: 10.1190/1.2213955
Wathelet M (2005) Array recordings of ambient vibrations: surface-wave inversion,
Doctor of Applied Science thesis, University of Liége. (available from:
[Link]
Weaver RL (2010) Equipartition and retrieval of Green’s function. Earthquake Science,
23(5), 397–402. doi: 10.1007/s11589-010-0738-2
Woodhouse JH (1974) Surface waves in a laterally varying layered structure. Geophys. J.
R. Astr. Soc., 37, 461–490. doi: 10.1111/j.1365-246X.1974.tb04098.x
Wuestefeld A, Wilks M (2019) How to twist and turn a fibre: Performance modelling for
optimum DAS acquisitions. The Leading Edge, 38, 226–231.
doI:10.1190/tle38030226.1
Wuestefeld A, Spica ZJ, Aderhold K, Huang H-S, Ma K-F, Lai VH, Miller MS, Urmantseva
L, Zapf D, Bowden DC, Edme P, Kiers T, Rinaldi AP, Tuinstra T, Jestin C, Diaz-Meza
S, Jousset P, Wollin C, Ugalde U, Barajas SR, Gaite B, Currenti G, Prestifilippo M,
Araki E, Tonegawa T, de Ridder S, Nowacki A, Lindner F, Schoenball M, Wetter C,
Zhu H-H, Baird AF, Rørstadbotnen RA, Ajo-Franklin J, Ma Y, Abbott RE,
Hodgkinson KM, Porritt RW, Stanciu C, Podrasky A, Hill D, Biondi B, Yuan S, Luo B,
Nikitin S, Morten JP, Dumitru V-A, Lienhart W, Cunningham E, Wang H (2023) The
Global DAS Month of February 2023. Seism. Res. Lett., 95(3), 1569–1577.
120 Bibliography

doi: 10.1785/0220230180
Yang J, Shragge J, Jin G (2022) Filtering strategies for deformation-rate Distributed
Acoustic Sensing. Sensors, 22, 8777. doi: 10.3390/s22228777
Yang Y, Atterholt JW, Shen Z, Muir JB, Williams EF, Zhan Z (2022) Sub-Kilometer
correlation between near-surface structure and ground motion measured with
distributed acoustic sensing. Geophys. Res. Lett., 49(1), e2021GL096503.
doi: 10.1029/2021GL096503
Yang Y, Zhan Z, Karrenbach M, Reid-McLaughlin A, Biondi E, Wiens, DA, Aster RC
(2024) Characterizing South Pole firn structure with fiber optic sensing. Geophys.
Rese. Lett., 51, e2024GL109183. doi: 10.1029/2024GL109183
Yu C, Zhan Z, Lindsey N J, Ajo-Franklin JB, Robertson M (2019) The potential of DAS in
teleseismic studies: Insights from the Goldstone experiment. Geophys. Res. Lett.,
46, 1320–1328. doi: 10.1029/2018GL081195
Yuan S, Lellouch A, Clapp RG, Biondi B (2020) Near-surface characterization using a
roadside distributed acoustic sensing array. The Leading Edge, 39(9), 646–653.
doi: 10.1190/tle39090646.1
Zeng X, Lancelle C, Thurber C, Fratta D, Wang H, Lord N, Chalari A, Clarke A (2017)
Properties of noise cross-correlation functions obtained from a distributed
acoustic sensing array at Garner Valley, California. Bull. Seism. Soc. Am, 107,
603–610. doi: 10.1785/0120160168
Zhan Z (2020) Distributed acoustic sensing turns fiber-optic cables into sensitive
seismic antennas. Seismol. Res. Lett., 91, 1–15. doi: 10.1785/0220190112
Zhang K, Li H, Wang X, Wang K (2020) Retrieval of shallow S-wave profiles from seismic
reflection surveying and traffic-induced noise. Geophysics, 85(6), EN105–EN117.
doi: 10.1190/geo2019-0845.1
Zhao Y, Li Y-E, Li B (2023) On beamforming of DAS ambient noise recorded in an
urban environment and Rayleigh-to-Love wave ratio estimation. J. Geophys. Res.
Solid Earth, 128, e2022JB026339. doi: 10.1029/2022JB026339
Zhu T, Shen J, Martin ER (2021) Sensing Earth and environment dynamics by
telecommunication fiber-optic sensors: An urban experiment in Pennsylvania,
USA. Solid Earth, 12, 219–235. doi: 10.5194/se-12-219-2021