ACTA UNIVERSITATIS
UPSALIENSIS UPPSALA
Digital Comprehensive Summaries of Uppsala Dissertations
from the Faculty of Science and Technology
1918
Source analysis of multiplet
earthquakes (two case studies in
Iran)
SAMAR AMINI
ISSN 1651-6214 ISBN 978-91-513-0909-5
Dissertation presented at Uppsala University to be publicly examined in Hambergsallen, Geocentrum, Villavägen 16, Uppsala, Friday, 12 June 2020 at 10:00 for the degree of Doctor of Philosophy. The examination will be conducted in English. Faculty examiner: Professor Kuvvet Atakan (University of Bergen, Norway).
Abstract
Amini, S. 2020. Source analysis of multiplet earthquakes (two case studies in Iran). Digital Comprehensive Summaries of Uppsala Dissertations from the Faculty of Science and Technology 1918. 45 pp. Uppsala: Acta Universitatis Upsaliensis. ISBN 978-91-513-0909-5. Multiplet earthquakes are large earthquakes of similar magnitude which occur close in time in the same limited geographical area. They are not common but they considerably increase the potential hazard in the area in which they occur. This thesis studies source properties and triggering mechanisms of two sets of multiplet events in Iran, which both occurred in unexpected areas, but close to some major active fault systems. The first multiplet is an earthquake doublet (Mw 6.5 and Mw 6.4) which occurred in northwestern Iran and caused more than 300 fatalities and significant injuries. In paper I, a teleseismic body-waveform inversion was used to deduce the slip distribution pattern on the fault plane of the first mainshock. The estimated slip pattern was utilized to calculate the Coulomb stress changes on the second fault plane and on the following aftershocks. Based on this analysis, the ambiguity between the primary and auxiliary fault plane of the second mainshock could be resolved. The second set of events is a triplet (Mw 6.1 - 6.0) that occurred in eastern Iran, close to the Kerman province. In paper II, the rupture propagation patterns of the three mainshocks were analyzed using Empirical Green’s Function (EGF) deconvolution. Two different approaches were used: One, the analysis of the azimuthal variation of the apparent rupture duration based on the width of the observed relative source time functions, and the second, the analysis of along-strike rupture directivity by assessing azimuthal variations of the relative amplitude spectra. The second approach was also used to investigate the rupture directivity of the largest aftershocks in the sequence (Mw 5 - 5.5). A clear tendency for rupture propagation towards the northwest was observed for the sequence, which suggests that the regional stress field has a central role in controlling the rupture propagation direction. In paper III, the slip distribution patterns of the triplet earthquakes were analyzed using teleseismic body-waveform inversion, and the similarities and differences in the rupture processes of the three mainshocks were investigated. Using the Coulomb stress analyses, the stress interactions between the mainshocks were examined, leading to identification of the primary and auxiliary planes. Finally, we suggest that the hazard estimates in complex continental regions such as Iran need to consider the probability of multiplets, which might allow a reduction of the seismic risk associated to the occurrence of further large earthquakes subsequent to a devastating earthquake. Keywords: Multiplet earthquakes, slip inversion, Coulomb stress, rupture directivity
Samar Amini, Department of Earth Sciences, Geophysics, Villav. 16, Uppsala University, SE-75236 Uppsala, Sweden.
© Samar Amini 2020 ISSN 1651-6214 ISBN 978-91-513-0909-5
Dedicated to those, who made this happen
Supervisor Roland Roberts
Proffessor at Department of Earth Sciences, Geophysics Uppsala University, Uppsala, Sweden
Assistant Supervisor Björn Lund
Associate Professor at Department of Earth Sciences, Geophysics Uppsala University, Uppsala, Sweden
Assistant Supervisor Hossein Shomali
Researcher at Department of Earth Sciences, Geophysics Uppsala University, Uppsala, Sweden
Opponent Kuvvet Atakan
Proffessor at Department of Earth Sciences University of Bergen, Bergen, Norway
List of papers
This thesis is based on the following papers, which are referred to in the text by their Roman numerals.
I Amini, S., Roberts, R., Raeesi, M., Shomali, Z.H., Lund, B., Zarifi, Z. (2018) Fault slip and identification of the second fault plane in the Varzeghan earthquake doublet. Journal of Seismology, 22, 815-831
II Amini, S., Roberts, R., Lund, B.(2020) Directivity analysis of the 2017 December Kerman earthquakes in Eastern Iran. Journal of Seismology, DOI: 10.1007/s10950-020-09913-8
III Amini, S., Raeesi, M., Roberts, R. (2020) Fault slip and rupture properties of the December 2017 Hojedk triplet in Eastern Iran. submitted to Geophysical Journal International
Reprints were made with permission from the publishers.
An additional journal article, published during my Ph.D. studies, that is not included in the thesis is:
Raeesi, M., Zarifi, Z., Nilfouroushan, F., Amini, S., Tiampo, K. (2017) Quantitative Analysis of Seismicity in Iran. Pure Appl. Geophys. 174, 793-833
Contents
1 Introduction . . . .9
2 Study area. . . .12
2.1 Seismotectonic setting of northwestern Iran and the doublet on August 11, 2012 . . . 13
2.2 Seismotectonic setting of eastern Iran and the triplet on December 2017 . . . 13
3 Methodology. . . 16
3.1 Waveform Modeling . . . 16
3.2 Coulomb stress . . . 19
3.3 Directivity analysis and Empirical Green’s Function . . . 20
4 Summary of papers. . . .23
4.1 Paper I: Fault slip and identification of the second fault plane in the Varzeghan earthquake doublet . . . 23
4.1.1 Motivation . . . 23
4.1.2 Results . . . 23
4.2 Paper II: Directivity analyses of the 2017 December Kerman earthquakes in Eastern Iran . . . 27
4.2.1 Motivation . . . 27
4.2.2 Results . . . 27
4.3 Paper III: Investigations of rupture properties of the December 2017 Hojedk triplet in Eastern Iran . . . .31
4.3.1 Motivation . . . 31 4.3.2 Results . . . 31 5 Concluding remarks . . . 36 6 Sammanfattning på svenska. . . 39 7 Acknowledgements . . . 41 References . . . .43
Abbreviations
3D Three-dimensional
EGF Empirical Green’s Function E-W East-West
g-CMT Global Centroid Moment Tensor
Hz Hertz
km kilometer
km/s kilometer per second Ml local Magnitude Mw moment Magnitude NE Northeast NW Northwest N-S North-South s second SE Southeast SW Southwest
IRIS Incorporated Research Institutions for Seismology IRSC Iranian Seismological Center
RSTF Relative-source-time-function STF Source-time-function
1. Introduction
Throughout human history some areas have been repeatedly devastated by earthquakes, the origins of which remained a mystery. As civilization and human constructions developed, understanding and mitigating the effects of these destructive events became ever more important. A major step towards the development of our understanding of earthquakes occurred in 1668 when Hooke introduced the theory of elasticity, explaining the deformation of solid objects due to external forces. Hooke’s law was extensively used to explain various aspects of the mechanical behavior of material and physical phenom-ena including earthquakes. Two and a half centuries later, the faulting the-ory of earthquakes was presented by Reid (1910) whose analysis was based on investigations of the San Andreas fault and the 1906 San Francisco quakes. Reid’s concept, known as the elastic rebound theory, explains earth-quakes as the sudden release of strain energy which has been accumulated slowly on two sides of a fault plane over an extended period of time. The theory was a major conceptual development in understanding the mechanisms of earthquakes. Further important developments towards understanding the large scale geological processes causing earthquakes came in the mid-1960’s after the installation of the World Wide Standardized Seismograph Network (WWSSN). The WWSSNallowed monitoring of earthquake activity all over the Earth with instruments that had similar, and well-known, characteristics. It was observed that most earthquakes occurred in long, narrow zones, later identified as boundaries between the tectonic plates. It is now understood that the surface of the Earth consists of a number of semi-rigid plates, which are in constant slow motion relative to each other, the motions being ultimately driven by the export of heat from the Earth’s hot interior. The idea that earth-quakes mainly reflect the relative motions of the tectonic plates was soon ex-tensively accepted.
About 90% of the total seismic energy around the world is released by earth-quakes that occur at plate boundaries, known as interplate earthearth-quakes. How-ever, some earthquakes occur within the interior of tectonic plates, away from the plate boundaries. These are known as intraplate regions. Intraplate earth-quakes can be catastrophic since they are rare and may happen in unexpected areas. They indicate that plate interiors are not fully rigid and include weak-ened zones where regional tectonic strain can be released. There are many populated areas with significant risk of large intraplate earthquakes, and it
is usually challenging to assess the seismic hazard in these areas, especially when there is a lack of historical earthquake records.
Globally, only a small extent of the large earthquakes belong to multiplet sets, but multiplets have been observed not infrequently on active tectonic boundaries, especially on subduction zones. Multiplets occur when adjacent plate segments or faults of comparable size break with a short time delay. One more specific definition of multiplets is that they are clusters of moderate to large earthquakes (M ≥ 6) with magnitude differences of no more than 0.2 units (e.g. Astiz and Kanamori, 1984), temporal separation of a small frac-tion of the average recurrence time of the earthquake cycle (Nomanbhoy and Ruff, 1996; Kagan and Jackson, 1999), and with spatial separation between the centroids of the earthquakes less than the rupture size of the events (Kagan and Jackson, 1999). Aftershocks following large earthquakes are known to be dangerous, because even relatively small aftershocks may cause the collapse of structures weakened by the preceding large event. Multiplets, where the following event or events are comparable in size to the first large event, are especially dangerous. It follows that better understanding of multiplets has the potential both to help us to understand earthquake systems and to improve hazard estimates and thereby reduce risk, possibly very significantly.
The proximity in space and time of multiplet events makes it clear that the first event has a significant role in triggering the following events, but the mechanisms of this triggering is not yet well-understood. However, statistical evidence suggests that the same physical triggering mechanism is responsi-ble for the occurrence of multiplets as for aftershocks and foreshocks (Felzer, 2004). A common theoretical framework for investigating the earthquake trig-gering process is to estimate stress interaction between the events. One of the most used methods is Coulomb stress transfer (Stein, 1999) which estimates the stress changes close to a fault caused by an earthquake’s fault displace-ment, and evaluates where these changes may enhance or suppress the likeli-hood of a subsequent event. The method has been successfully used to explain the triggering mechanisms of several successive mainshocks (e.g. King et al., 1994) and multiplets (e.g. Lin et al., 2008).
In this thesis, two multiplet sequences are discussed from the point of view of the slip distribution and rupture process. The events investigated occurred in Iran, which is an active continental deformation zone with significant levels of seismic hazard and where numerous destructive earthquakes are known to have occurred in both historical and recent times.
This dissertation consists of five chapters including this introduction as the first chapter. Chapter two describes the tectonic and seismicity characteristics of the areas studied. A concise description of the methods used for the
follow-ing investigations is presented in chapter three. Chapter four is a summary of the papers included in this thesis and in chapter five conclusions and outlook are discussed. Following this, the papers are included in the printed versions and pre-prints.
In paper I, entitled ”Fault slip and identification of the second fault plane in the Varzeghan earthquake doublet”, a teleseismic body-waveform inversion method was used to deduce the slip distribution for the first mainshock of a doublet in 2012. The deduced slip pattern was then used to estimate the Coulomb stress changes on the two nodal planes of the second mainshock and the largest aftershocks.
In paper II, ”Directivity analysis of the 2017 December Kerman earth-quakes in Eastern Iran”, the rupture propagation of the Kerman triplet was investigated, utilizing Empirical Greens Function (EGF) deconvolution. Two different approaches were applied separately to P and S phases to detect the rupture propagation direction for the three mainshocks and six of the largest aftershocks.
In paper III, ”Investigations of rupture properties of the December 2017 Hojedk triplet in Eastern Iran”, the fault plane parameters and the slip distri-bution pattern of the three mainshocks were investigated using a teleseismic body-waveform inversion method. The estimated slip models were used to as-sess the Coulomb stress changes on the fault planes and to examine the stress interaction between them.
2. Study area
Iran is part of the enormous Alpine-Himalayan orogenic belt. Active conver-gence of the Arabian plate to the southeast and the Eurasian plate to the north-west of Iran cause intense seismicity, taking place on active thrust and strike-slip faults (e.g. Jackson, 1992) (Figure 2.1). The crustal deformation resulting from the continental collision has generated various tectonic and topographic features in a relatively small area, dividing Iran into five major seismotectonic provinces (Mirzaei et al., 1998).
Figure 2.1. Tectonic map of the Arabian-Eurasian collision. Focal mechanisms are from the global-CMT catalog for the period from 1976 to 2020. Major faults are in red. Arrows and the numbers beside are the GPS-derived plate velocities (mm/year) rela-tive to Eurasia according to Reilinger et al. (2006). Inset shows the Alpine-Himalayan Orogenic belt. Location of other figures is marked by green boxes.
Iran have experienced frequent destructive earthquakes back in historical times and during recent decades which totally ruined several populated cities
and caused large numbers of fatalities and injuries (e.g. the 1990 Mw 7.4 Rudbar and 2003 Mw 6.6 Bam earthquakes). In the research described below, two seismic sequences in two different seismogenic zones are investigated; an earthquake doublet in northwestern Iran and a triplet in eastern Iran.
2.1 Seismotectonic setting of northwestern Iran and the
doublet on August 11, 2012
The most dominant tectonic element of northwestern Iran is the North Tabriz fault which is a large strike-slip fault with NW-SE strike. With an overall length of more than 200 km along several right-stepping en-echelon segments, it accommodates about 7 mm/yr of right-lateral motion of the Arabia-Eurasia convergence (e.g. Masson et al., 2006). The North Tabriz fault has experi-enced a significant number of historical earthquakes leading to repeated de-struction of Tabriz city, which now has a population of over 2 million. How-ever, recent seismicity data from 1960 to 10 August 2012 from EHB Bulletin (http://www.isc.ac.uk) do not include any earthquake of larger than magnitude 5 and the most recent major movement was a sequence in 1721-1786 (M 6.3 to 7.3) (Figure 2.2). According to the slip accumulation rate and historical seismicity data, the average recurrence time of a magnitude 7-7.4 earthquake on the North Tabriz fault is 250-300 years which suggests a high risk of a sig-nificant earthquake in the coming decades. The Varzeghan doublet, occurred about 50 km north of the North Tabriz fault in the area where there are no reports of previous seismicity. The sequence caused about 300 fatalities and more than 3000 injuries.
2.2 Seismotectonic setting of eastern Iran and the triplet
on December 2017
Eastern Iran is an almost flat, rigid block, which is surrounded by large strike-slip faults to the east, west and north. The triplet occurred on the western mar-gin of this rigid block and is surrounded by several major right-lateral strike slip faults and some smaller reverse faults (Figure 2.3). The recent seismic ac-tivity is mostly concentrated on the Gowk fault to the south, hosting destruc-tive earthquakes in 1981(Mw 7.1 Sirch, Mw 6.6 Golbaf) and 1998 (Mw 6.6 Fandoqa) (Berberian et al., 2001). The Gowk fault with a length of ∼ 150 km and a strike of ∼ 155◦, breaks into the Nayband, Lakar-Kuh, and Kuh-Banan
Figure 2.2.Seismicity of northwestern Iran for M ≥ 3 events based on the ISC catalog for the period 1960 to 10 August 2012. Stars and beach balls (from the global-CMT catalog) mark the Ahar-Varzeghan doublet. The North Tabriz Fault is plotted in red. Green ellipses are the last four historical events based on Berberian and Yeats (1999). Gray boxes are main cities.
faults to the north. The first two faults striking N-S with right-lateral shear motion have experienced relatively small seismicity, but the third fault has a long record of earthquakes back in 1875 (e.g. Ambraseys, N.N. and Melville, C.P., 1982). The Kuh-banan fault extending for ∼ 200 km along NW-SE was lately ruptured in 1977 (Mw 5.8 Gisk) and 2005 (Mw 6.4 Zarand). While the first event indicated an almost pure strike slip motion (Berberian et al., 1979), the latter one had a pure reverse mechanism on an E-W trending, north dipping splay fault which was assumed to be the southern termination of the Kuh-banan fault. The 2017 triplet occurred 40 km east of the 2005 event with all mainshocks presenting oblique reverse movement. Further investigations of faulting in the area (Walker et al., 2010), have detected a series of short strike-slip and reverse faults along the margins and even within the mountain-ous regions which suggests evolution of a ∼ 40 km wide left step restraining bend between the Gowk and Kuh-Banan faults.
1977
2005
1998 1981
1989
Figure 2.3. Seismicity map of the triplet area. All focal mechanisms are from the global-CMT catalog. Blue circles are events with M ≥ 5 between 1964 to Nov 2017, from the ISC catalog. Stars and colored focal mechanisms are the 2017 triplet events. Faults are in red. White boxes are main cities.
3. Methodology
Earthquakes radiate mechanical (seismic) waves which propagate from the earthquake in all directions through the Earth’s interior and are recorded by sensitive instruments placed on or close to the surface. The record of the motion, called a seismogram, includes information about both the source and the medium it passed through. In a general form the observed seismogram can be written as:
u(t) = s(t) ∗ g(t) ∗ i(t) (3.1)
where u(t) is the seismogram, * is the convolution operator, s(t) is the seis-mic source signals, g(t) is the propagation effect, and i(t) is the instrument re-sponse of the seismograph. It is possible to study a specific component of the above equation if this effect can be isolated from the effects of the other com-ponents. Approaches to extracting information on the seismic source (s(t)) by use of waveform inversion and Empirical Green’s function methods are going to be discussed in this thesis.
3.1 Waveform Modeling
Waveform modeling is an iterative process of comparing synthetic and ob-served seismograms to minimize the difference between them by adopting the propagation structure or the seismic source process. The propagation effect (g(t) in Eq. 3.1), also known as the Earth transfer function or the Green’s function, is the most complex parameter since it needs to account for elastic phenomena, such as reflections, multiplications, diffractions and phase con-versions, as well as attenuation effect, including scattering and geometrical spreading. However, at teleseismic epicentral distances (30◦− 90◦), we can reasonably avoid the crustal multiplications and core-mantle boundary diffrac-tions. Then, considering the early P-wave arrivals, g(t) can be regarded as basically consisting of three pulses; P, pP and sP, i.e. the P wave which has traveled directly from source to receiver, and the signals which have first prop-agated to the surface as P or S waves before propagating onwards as P. Re-flected and phase-converted pulses will arrive at the receiver after the direct
(primary) waves. Arrival time differences mainly depend on source depth and Earth structure and the relative amplitudes depend primarily on radiation pat-tern and epicentral distance (Lay and Wallace, 1995, chapter 10).
Seismic source parameters generally include fault orientation, epicenter, depth, seismic moment, slip direction and magnitude, and source time function (STF). As mentioned above, the Green’s functions produce synthetic wave-forms that traveled along the path from the source to the seismometer. Hence, the seismic moment and source time function can be estimated by compari-son of the observed and synthetic waveform amplitudes. Once the source time function is determined, the other source parameters may be estimated under some assumptions, including an assumed rupture velocity. Unfortunately, such analyses are often not so straightforward because there is a trade-off between the assessed parameters. The strongest trade-off is mostly found between the source depth and duration of the STF (Lay and Wallace, 1995, chapter 10), where the latter is proportional to the rupture velocity. In the other words, a deeper source with higher rupture velocity would be similar to a shallower source and lower rupture velocity. However, use of broadband data and mul-tiple stations for modeling is helpful to reduce the trade-offs (Christensen and Ruff, 1985).
In the following, the source parameters and slip distribution are estimated by inversion of body waves at teleseismic distances using the computer code of (Kikuchi et al., 1993). In this method the fault plane is discretized into a number of elements or sub-faults with equal dimensions, covering the fault, which has predefined strike and dip angles. The kinematic parameters, slip and source time function, are retrieved for each sub-fault by adjusting the nu-merical model of these to obtain a good fit between the observed seismograms and the model response. The far-field displacement uj(t) at station j due to
shear dislocation on a fault surface S can be represented as (Olson and Apsel, 1982; Hartzell and Heaton, 1983):
uj(t) = 2
∑
q=1 Z Gq j(t, ξ ) ∗ ˙Dq(t, ξ ) dξ (3.2)where * is the convolution in the time domain, Gq j are the synthetic
dis-placement waveforms (Green’s functions) for a source at ξ and receiver at station j computed for two orthogonal unit steps q. ˙Dqis the q − th
compo-nent of spatio-temporal slip rate which can be written as a linear combination of space and time basis functions:
˙ Dq(t, ξ ) ' K
∑
k=1 L∑
l=1 aqklXK(ξ )Tl(t − tK) (3.3)with k and l as space and time steps, aqkl is an array of coefficients that
needs to be determined, XK(ξ ) is the space basis function at grid node k, Tl(t −
tK) is the time basis function where tkis the time that rupture starts at grid node
k. By substituting Eq. 3.3 in Eq. 3.2:
uj(t) = 2
∑
q=1 K∑
k=1 L∑
l=1 aqklTl(t − tK) ∗ Z Xk(ξ ) Gq, j(t, ξ ) dξ (3.4)We can write the time and space functions as a single parameter Hqkl j and
substitute in Eq. 3.4 to:
uj(t) = 2
∑
q=1 K∑
k=1 L∑
l=1 Hqkl jaqkl (3.5)In vector form, Eq. 3.5 can be written as u = Ha. Solving for a using a non-negative least squares inversion method retrieves and estimates slip on each fault element, which is zero or positive. The least squares misfit defined as ∑(xobs− xcal)2/ ∑(xobs)2 is minimized during the inversion. xobsis the
ob-served data and xcalis the model data.
To select a model among a group of possible models the Akaike Bayesian Information Criterion (ABIC; Sclove, 1987)was used here. The Bayesian pro-cess chooses two weights for the orthogonal Green’s functions and one for the
STFduration, when evaluating the goodness of fit between the observed and modeled data. The two weights that scale the orthogonal Green’s functions define the slip amplitude and rake angle, where the rake angle is calculated at ±45◦ from the predefined average rake. The STF is described by over-lapping triangles in which the amplitudes of the triangles are parameters in the inversion. A smoothness constraint is applied as prior information and a constant rupture velocity is assumed during the inversion. Theoretically, by solving an inverse problem, the kinematic parameters can be retrieved at each point of the fault. One problem with this type of analysis is the high degree of non-uniqueness in the solution (Das and Kostrov, 1994), in the sense that different models may fit the data almost equally well. Therefore, other types of information such as the existence of a surface rupture, aftershock
distribu-tions, geological information, or information from other types of geophysical analysis, may be very helpful in stabilizing the results.
3.2 Coulomb stress
A sudden release of the accumulated shear stress is the cause of most natu-ral earthquakes. This stress release causes changes of the shear and normal stresses in the volume surrounding the source. Such a stress change can sig-nificantly affect the seismicity rate, provoke earthquake sequences, clustering and aftershocks (e.g. Harris, 1998; Stein, 1999). One of the most known meth-ods for studying earthquake interactions is the Coulomb stress change:
∆σf = ∆σs+ ´µ ∆σn (3.6)
The Coulomb stress change (∆σf), also sometimes known as the Coulomb
failure stress (CFS), considers the changes in shear stress (∆σs) and normal
stress (∆σn) on a chosen or predefined ”receiver” fault with an effective
fric-tion coefficient ( ´µ ). Shear stress is defined as positive in the direction of fault slip and the normal stress as positive for opening (unclamping). Theoretically, failure is encouraged where the Coulomb stress change is positive and is dis-couraged for negative values.
To evaluate the static stress changes caused by a significant earthquake, we use the Coulomb3 software (Toda et al., 2011). The slip distribution patterns obtained by waveform inversion were used as input for the calculation. The code uses the slip values to calculate the 3D strain field, multiplies this by elastic stiffness, and produces the stress changes due to the source fault plane (the fault that has ruptured). The stress changes are then used to estimate the shear and normal stress changes on specified receiver fault planes. Each receiver fault plane is defined by its position, strike, dip and rake values. The shear stress changes are sensitive to the position of the receiver fault relative to the source fault and the relative geometries (rakes and strikes) of the faults. The normal stress changes are affected by the geometry and position, but not by the rake of the receiver fault.
An earthquake can thus enhance or suppress the likelihood of subsequent events, depending on their location and orientation. There are some implicit assumptions about the pre-existing stress situation before the main event, but numerous studies have proved this type of modeling to be a useful analysis
tool (e.g. King et al., 1994; Stein, 1999). The seismic hazard in an area may be significantly affected by stress transfer from earlier earthquakes. However, the Coulomb stress criterion only indicates if the receiver fault has been brought closer to failure or not, and nothing about how close it is. Evaluation of the latter demands accurate assessment of the pre-existing stress fields.
3.3 Directivity analysis and Empirical Green’s Function
A common representation of the seismic source signal is the source time func-tion, which is the time evolution of the seismic moment. TheSTFconsists of the slip history (TD) and rupture time (TR). Slip during an earthquake is often
assumed to increase gradually on the fault plane and may be modeled as a ramp function where the derivative is a box car with its length equal to the slip duration (e.g. Lay and Wallace, 1995, chapter 9). The rupture time describes the rupture duration along the finite fault. Considering that the waves radiated from the initial part of the rupture would propagate first and then the points further along the fault, the rupture-function also may be assumed to be well described by a box car. The convolution of these two box cars gives a trape-zoid with its area equal to the seismic moment (Figure 3.1a). If we assume a finite fault of length L, and a simple pulse type rupture model (e.g. Haskell, 1964) propagating from one end through the other end of the fault with rupture velocity vr, and a receiver recording the rupture at distance r0and azimuth θ
(Figure 3.1.b), then the variation of the rupture duration with azimuth (Tθ) can
be written as: Tθ = L vr 1 −vr c cos θ (3.7)
According to Eq. 3.7, unilateral rupture, would manifest itself as an appar-ently shorter durationSTFwhen observed at stations in the direction of rupture propagation, and an apparently longer duration when observed from the other side. Since the area of the source time function is constant (because M0 is
constant) shorter duration is accompanied by higher amplitude observedSTF
signals, and vice versa (Figure 3.1.c) (after Stein and Wysession, 2003).
As mentioned above, to extract the source properties from the seismograms we need to filter out the wave propagation effects as well as site and instru-mental effects. A practical method to suppress these effects is the Empirical Green’s Function (EGF) method (Hartzell, 1978; Mueller, 1985). The method involves deconvolution of the target event with smaller events which are
suffi-Figure 3.1. (a) Convolution of tow box-car functions representing the slip history (TD) and the rupture time (TR) results in a trapezoid source time function. (b) A finite
fault of length L and a recording station at distance r0and azimuth θ from the rupture
initiation point. Rupture propagates with rupture velocity vrfrom one end of the fault
to the other.(c) Variation of the source time function at different azimuths as an effect of rupture directivity (after Stein and Wysession, 2003)
ciently close to the target event such that the propagation paths can be assumed to be essentially similar. The smaller event should also have a similar fault-ing mechanism and be at least one order of magnitude smaller than the target event, so that it can be regarded as a point source. The deconvolution provides a relative source time function (RSTF), which contains information on source properties such as rupture extent and propagation direction.
In paper II, we applied theEGFmethod to capture the variation of rupture duration with respect to azimuth and so to detect potential rupture directiv-ity. Two different approaches were used in order to validate the results. One approach was to retrieve the RSTF and assess the variations of its width at different stations. The other approach was the comparison of the displace-ment amplitude spectrum observed at the stations along and opposite to the fault strike. The latter approach stems from the fact that the corner frequency, defining a change of gradient in the observed source spectrum, is inversely proportional to the source duration ( fc= 1/TR+ 2/TD). Hence, the amplitude
spectrum will exhibit an azimuthally dependent corner frequency (Figure 3.2) (Calderoni et al., 2013, 2015). By averaging the spectra at stations aligned along fault strike and comparing with the average of stations in the opposite direction, one can detect the propagation direction of the rupture.
Figure 3.2. Illustration of the along strike rupture directivity based on the spectral variation at stations at opposite azimuths (Calderoni et al., 2015),
4. Summary of papers
4.1 Paper I: Fault slip and identification of the second
fault plane in the Varzeghan earthquake doublet
4.1.1 Motivation
On August 11, 2012, northwestern Iran experienced two destructive earth-quakes with Mw 6.5 (at 12:23 GMT) and Mw 6.4 (at 12:34 GMT) followed by numerous aftershocks with magnitudes up to Mw 5.6. According to the global-CMT(g-CMT) solution, the first mainshock ruptured with almost pure strike-slip motion, and the second main shock with an oblique thrust mech-anism. Due to the short time delay between the mainshocks, and the small epicentral separation between them, ∼ 6 km (Ghods et al., 2015), we consider them as doublet.
A surface rupture trace of about 12 km length with an E-W trend has been tracked in the field. The trace was associated to the first mainshock, since it was favorably oriented along one of its nodal planes (Donner et al., 2015). The second mainshock has no associated surface rupture, nor a clear aftershock signature to identify the fault plane. There is also no recognized active fault trace in the area where the doublet occurred and an absence of significant seismic activity for at least the past 200 years. However, the doublet locates about 50 km northeast of the North Tabriz Fault, a large active strike-slip fault (Figure 2.2) with high risk of a significant earthquake in the coming decades. Therefore seismic investigations of the doublet have important implications for hazard assessment in the area.
4.1.2 Results
A teleseismic body waveform inversion method (described in section 3.1) was used to estimate the slip distribution pattern associated with the first main-shock. The data for the second mainshock was highly contaminated by the surface waves and coda from the first event, thus preventing reliable results
from body wave inversion with this data for the event. P and SH waveforms from 93 stations at teleseismic distances, obtained from the IRIS data center (http://www.iris.edu/), were used for the inversion. A regional velocity model and the CRUST2 model (Bassin et al., 2000) were chosen to describe the near source and near receiver structures, respectively. A zero-phase band-pass fil-ter of 0.015 to 0.5 Hz was applied to the data to enhance the body waves. The fault plane was discretized with 4 km node spacing, a total of 32 km length along strike, and 20 km width along dip direction. The initial strike and dip values were taken from the g-CMTsolution and then altered gradually to obtain a minimum misfit between the observed seismograms and the model response. The hypocenter depth and the rupture velocity was controlled man-ually to provide a good match with the observed surface rupture as well as a good fit between the observed and synthetic data.
Figure 4.1.The estimated slip distribution pattern for the first mainshock of the 2012 Varzeghan doublet. Horizontal axis is distance along strike and vertical axis is dis-tance along dip of the fault planes. Yellow star is the hypocenter. Red dots are the aftershocks with magnitude larger than 3. The topography profile along the strike of the fault is plotted above the slip map, and the red triangles on it marks the two ends of the observed surface rupture.
We conducted separate and joint inversions of P and SH data and examined various combinations of predefined parameters. The preferred slip pattern ob-tained for the first event, resolves surface slip vectors which matches reason-ably well with the observed surface rupture (Figure 4.1). The slip model with a hypocenter depth at 12km and rupture velocity of 2.8 km/s, yields a slip
distribution pattern that extends more than 30 km along the strike and 20 km along the dip of the fault plane. The maximum slips occur around 4 km depth and to the west of the hypocenter. The rupture propagation directionality is mainly from the east to the west.
Figure 4.2. Coulomb stress changes caused by the first mainshock on the two nodal planes of the second mainshock, (a) North-dipping nodal plane, (b) East-dipping nodal plane. (c), (d) The Coulomb stress changes on nodal planes of the largest aftershocks, due to the slip on the first event combined with the north-dipping and east-dipping nodal planes of the second event, respectively. Star is the second mainshock, circles are the ten largest aftershocks. F1 and F2 are the surface traces of the fault planes associated to the first and second mainshocks.
The slip model obtained for the first event was used as the source fault to calculate the Coulomb stress changes on the two possible fault planes of the second mainshock as the receivers (section 3.2). We compared the deduced Coulomb stress changes on both these nodal planes (Figure 4.2) to investigate which was the more probable fault plane. The fault plane solutions (strike, dip, rake) and the hypocenter depth for the second event were defined according to the g-CMTsolution and the relocations of Ghods et al. (2015), respectively. The sections of the Coulomb stress along these two planes indicate that the Coulomb stress changes are almost neutral around the hypocenter for the E-W nodal plane (Figure 4.2a), while they are strongly positive on the N-S nodal plane (Figure 4.2b).
Furthermore, we investigated Coulomb stress changes caused by both main-shocks on the nodal planes of the ten largest aftermain-shocks. For this analysis we defined a simple slip model for the second event based on its focal mechanism solution and scalar moment (from g-CMT). Nine out of ten of the aftershocks exhibited positive Coulomb stress changes when the N-S nodal plane of the second mainshock was combined with the first fault plane, but only half were positive when the E-W plane was used instead (Figure 4.2 d, c).
4.2 Paper II: Directivity analyses of the 2017 December
Kerman earthquakes in Eastern Iran
4.2.1 Motivation
Three moderately large events with Mw 6.1-6.0 occurred in eastern Iran be-tween December 1st and 12th 2017. The first two events are estimated to have 3 km epicentral separation (based on the IRSC catalog), 11 days time separa-tion, and very similar focal mechanism solutions (based on the g-CMT). The third event is located about 10 km NW of the first two, it occurred 13 hours after the second one, and had a slightly different mechanism. The sequence was followed by seven moderate aftershocks with Ml 5.0-5.2 and a significant number of smaller events. We studied the rupture propagation direction of the mainshocks and six of the largest aftershocks in the sequence, and compared the results with the regional tectonic stress state.
4.2.2 Results
In order to detect the rupture directivity of the events in the triplet, we applied the EGF approach (section 3.3) using P and S wave data separately. Three
EGF events were selected for each mainshock analysis (EGF1 to EGF9 re-spectively), to reduce the risk of radiation pattern mismatch between the target andEGFevent. The P-waves were used to retrieve theRSTFand compare its width at different stations. For this approach, first a 10 s window bracketing the maximum P-wave amplitude was selected on vertical seismograms of both the mainshock and EGFwaveforms, which were all band-pass filtered 0.5 to 1.2 Hz before the selection. Then, a 3 s window of theEGF waveforms was chosen at each station in such a way as to include the most similar part of the P-waveforms. Next, the target and theEGFevent were deconvolved with a 1% water-level regularization. The resulting RSTF waveforms have been plotted according to the stations’ azimuth to assess the relative duration of the observed pulses.
The S-wave data was used in a different approach, to reach an additional semi-independent analysis allowing assessment of the consistency of the re-sults. 30 s data segments including a significant part of the S-wave energy, were chosen on both radial and transverse S-waveforms of the mainshock and
EGFevents. Amplitude spectra of the waveforms were calculated and spectral division of the mainshocks and EGF’swas performed. Then stations aligned
within ±45◦ with fault strike were chosen and used to calculate the average spectra along fault strike.
Figure 4.3. (a) Retrieved width of the RSTF for the first mainshock. The waveforms are normalized (dotted plots). Solid plots are the absolute values of the waveforms and gray curve draws an envelope over it. Dash-dotted vertical line marks the beginning of the RSTF and triangles mark the ends, both determined based on 35% drop from the peak value of the envelope. (b), (c) Results of the S-wave spectral division for the first mainshock using two different EGF events, EGF1 and EGF3, respectively. Red is the average spectra of the stations around the NW end of the fault plane, blue is the average around the SE end, and black is the average for the rest of the stations. The bars in the insets show the mean amount of the amplitude spectra in the frequency range of 0.5 to 1.2 Hz, using the same color scheme as the spectra curves.
Our directivity analyses for the three mainshocks are summarized in Figures 4.3, 4.4, and 4.5. The left columns of each figure displays theRSTF calcula-tions where a shorter width at stacalcula-tions oriented toward one direction would suggest a rupture directivity in this direction. The column to the right shows the S-wave analysis, where average spectra over stations in the NW and SE directions of the fault strike are plotted in red and blue, respectively, and av-erage spectra of the stations in the directions perpendicular to the fault strike are plotted in gray. The largest average spectra should define the prevailing rupture propagation direction, but if the gray spectra has greater amplitude, the test is considered as unreliable. For a more convenient comparison, the av-erage spectral values are summed in a band width of 0.5 to 1.2 Hz, and plotted
as bar graphs shown as the inset in the spectral diagrams. The choice of fre-quency band was constrained to be adequately high relative to the frefre-quency content of the mainshocks, and sufficiently low for theEGF events. The latter is such that the EGFs can effectively be considered as point sources, and the former is to capture the finite fault properties of the mainshocks.
Figure 4.4. (a) Retrieved width of the RSTF for the second mainshock. (b), (c), (d) Results of the S-wave spectral division for the second mainshock using three different EGF events, EGF5, EGF6 and EGF7, respectively. Details as in Figure 4.3.
For the first mainshock, the retrieved width of the RSTF shows a relatively shorter duration for stations oriented towards the NW (Figure 4.3a). With the spectral analyses, the test was assessed to be invalid usingEGF1, due to the higher gray curve (Figure 4.3b). The S-wave data was corrupt at some sta-tions for EGF2 and analysis with this event was not feasible, but withEGF3 the results suggest directivity towards the NW (Figure 4.3c). For the second mainshock, not only the RSTF width comparison suggests rupture directiv-ity toward NW (Figure 4.4a), but also the spectral division analysis all show a clear rupture propagation direction to the NW direction (Figure 4.4b,c,d). For the third mainshock, the waveform deconvolution results do not exhibit a clear rupture propagation direction (Figure 4.5a). Besides, the spectral analy-ses display an almost uniform rupture propagation usingEGF7 (Figure 4.5b),
and apparent rupture directivity toward SE and NW using EGF8 andEGF9, respectively (Figure 4.5c,d). This discrepancy of the obtained results could suggest the lack of significant along strike directivity, or a bilateral rupture propagation.
Figure 4.5. (a) Retrieved width of the RSTF for the third mainshock. (b), (c), (d) Results of the S-wave spectral division for the third mainshock using three different EGF events, EGF5, EGF6 and EGF7, respectively. Details as in Figure 4.3.
Using S-wave spectral division analysis, six of the largest aftershocks of the sequence with Ml 5.0 to 5.2, have been analyzed to detect any preferred rupture propagation direction which four of them showed rupture propaga-tion towards the northwest. Our directivity analysis suggests a predominant rupture propagation direction towards the NW for the seismic sequence. The directionality seems to be driven by the regional tectonic stress field due to the northward Arabian-Eurasian convergence.
4.3 Paper III: Investigations of rupture properties of the
December 2017 Hojedk triplet in Eastern Iran
4.3.1 Motivation
The 2017 seismic sequence in eastern Iran consisted of three mainshocks with Mw 6.1-6.0 and numerous aftershocks with Mw up to 5.2. The sequence oc-curred in a sparsely-populated area and caused minor damage, but it was just 60 km north of Kerman city, with a population of almost one million and sev-eral earlier destructive earthquake incidents. The triplet area is surrounded by four major right lateral strike slip faults: one to the southeast, two to the north, and one to the northwest (Figure 2.3). The triplet is located in a gap between these faults and rupturing was predominantly reverse faulting, incon-sistent with the mechanisms observed for the surrounding faults. The faulting patterns and field observations in the area suggest a wide restraining bend connecting these strike-slip faults together by creating a series of short reverse faults (Walker et al., 2010). The 2017 seismic sequence took place within the restraining bend and at the eastern edge of it. Therefore, investigating source properties of the three mainshocks is helpful to better understand the faulting mechanism and stress interactions in a restraining bend.
4.3.2 Results
To estimate the slip distribution pattern on the fault plane of the three main-shocks, we applied the waveform inversion method by Kikuchi et al. (1993), described in section 3.1. The waveform data was from 40 stations at tele-seismic distances, obtained from the IRIS data center. After removal of the instrument responses and bandpass filtering 0.01 – 1 Hz, a window of 25 s length starting 3 s before the first P and S arrivals was used for the inver-sion. The near-source and receiver structures were defined according to the Crust2 (Bassin et al., 2000) and Iaspei91 (Kennett and Engdahl, 1991) veloc-ity models, respectively. The initial strike and dip angles of the faults were defined based on the g-CMTsolution and the hypocenter depths and locations are from the IRSC catalog. These predefined values are then adjusted man-ually and iteratively to fulfill four criteria including reduction of the misfit value, consistency with the observed surface rupture, compatibility of the es-timated moment magnitude with the g-CMTmoment, and smoothness of the slip model. The first two mainshocks did not have any geological indications or field observation of rupture that could be applied to deduce which of the two nodal planes was the fault plane. Therefore, we estimated the slip
distri-Table 4.1. Fault plane parameters for the slip models presented in Figure 4.6.
Model strike,dip,rake depth(km), vr(km/s) M0(e16), misfit strike,dip,rake depth(km), vr(km/s) M0(e16), misfit rupture duration a, b 305◦, 52◦, 103◦ 8, 2.0 165, 0.26 105◦, 40◦, 70◦ 8, 1.9 171, 0.26 6 s c, d 310◦, 50◦, 99◦ 9, 1.8 123, 0.39 115◦, 40◦, 81◦ 8, 1.6 134, 0.38 12 s e 110◦, 40◦, 85◦ 6, 2.2 143, 0.40 8 s
bution on both nodal planes of the first two mainshocks. The third mainshock however, was observed to have surface rupture and a SW-dipping nodal plane was suggested as the fault plane (Savidge et al., 2019).
The estimated slip distribution patterns and the fault parameters are dis-played in Figure 4.6 and Table 4.1. Comparing the results, we observe that the fault strike, dip, rake, hypocenter depth and rupture velocity are quite similar for the first two mainshocks (Table 4.1), but the rupture propagation pattern is different. For the first mainshock, the rupture propagation is almost bilateral with only slight further extension to the northwest (Figure 4.6a, b), but for the second mainshock, slip vectors show a predominant rupture propagation direction towards the northwest (Figure 4.6c, d). The third mainshock appears different from the first two in several ways; shallower hypocenter, 6 km com-pared to 8 - 9 km, faster rupture velocity, 2.2 km/s comcom-pared to 1.8 - 2.0 km/s, a clear bilateral rupture propagation, and the surface rupture which was absent in the first two events.
The estimated slip models have been used to assess the Coulomb stress changes on the faults in two steps. First, the slip models for the first main-shock were defined as the source and the Coulomb stress changes calculated on the possible fault planes of the second mainshock. For an easier visual comparison, the contour lines of the slip pattern for the receiver fault is over-lain on the calculated Coulomb stress changes and the results presented for the four possible combinations of the nodal planes (Figure 4.7a, b, c, d). Compar-ing the correlation between the areas with increased Coulomb stress changes (red cells) and the overlain slip distribution pattern, the better correlation is observed for the source fault as the SW-dipping nodal plane of the first main-shock and the receiver fault as the NE-dipping plane of the second mainmain-shock (Figure 4.7c).
In the next step, Coulomb stress was calculated on the fault plane of the third mainshock, due to the rupture of the first two mainshocks, considering the four combination of the nodal planes (Figure 4.8a, b, c, d). As above, the increased Coulomb stress changes are most compatible with the slip distribu-tion pattern when the SW-dipping plane of the first mainshock and the
NE-Figure 4.6. Slip distribution pattern estimated for (a), (b) The first mainshock of the Kerman triplet, (c), (d) the second mainshock, and (d) the third mainshock. In the right column are the solution for SW-dipping nodal planes and to the left are the NE-dipping ones. Hypocenter is at node (0,0) and is marked with a star. Numerical parameters of the models are listed in Table 4.1.
dipping plane of the second mainshock are combined together (Figure 4.8c). Therefore, we suggest these two planes as the actual fault planes that have ruptured during the mainshocks.
Figure 4.7. Coulomb stress changes due to the first mainshock on the fault plane of the second mainshock. Four possible combinations of the nodal planes are presented. The second fault plane (receiver fault) is outlined by the green frame and the contour lines of the slip pattern obtained by the waveform inversion is overlain on it. Gray star marks the hypocenter. Fault labels are ”F1” as the first mainshock and ”F2” as the second mainshock. ”-NE” means the NE-dipping nodal plane and ”-SW” means SW-dipping. Northwest and southeast directions along the receiver fault strike are indicated by the NW and SE labels.
Figure 4.8. Coulomb stress changes due to the first and second mainshocks on the fault plane of the third mainshock. Four possible combination of the source faults are presented. The third fault plane (receiver fault) is outlined by the magenta frame and the contour lines of slip pattern obtained by the waveform inversion is overlain. Gray star marks the hypocenter. Fault label ”F3” stands for the third mainshock. Other labels as in Figure 4.7.
5. Concluding remarks
In this research we studied source properties and triggering mechanisms of two multiplet sequences in Iran, which both happened in unexpected areas but close to major active fault systems. The first event was the Ahar earthquake doublet in 2012 (Mw 6.5 and Mw 6.4) that occurred in northwestern Iran. Before the doublet, the knowledge of active faulting in the region was very limited and it was thought that the North Tabriz Fault was accommodating all the shortening in northwestern Iran (e.g. Masson et al., 2006), with the area north of it being an effectively rigid block (e.g. Djamour et al., 2011). This was consistent with the seismic records until 2012, but, then the Ahar doublet occurred 50 km north of the North Tabriz Fault, with a different strike and sense of motion. After the doublet, studies suggested that the shear strain in the area is not fully compensated by the North Tabriz Fault and the remaining shear strain is transferred further north and accommodated during the 2012 earthquake sequence (e.g. Donner et al., 2015). The occurrence of the 2012 doublet provided an opportunity to improve our knowledge of active faults in the region. It also provided more general insights into the seismic hazard of areas adjacent to continental active faults, which are usually assumed to be almost rigid blocks with relatively little internal deformation.
The first mainshock in the 2012 doublet sequence ruptured with a strike slip mechanism on an E-W trending fault plane. The second mainshock was characterized by oblique thrust motion, but the most important issue to be ad-dressed for this event was to identify the fault plane to determine if it was striking E-W, similar to the first event, or if it was perpendicular to that. Us-ing Coulomb stress analysis we suggested that the fault plane of the second mainshock was the NNE-SSW striking plane, which intersects with the first fault plane at depth. Fig 5.1a displays 3D illustration of the fault planes where the intersection is visible. Moreover, our results for the slip model of the first mainshock resolved two distinct slip patches which have different pre-dominant slip directions. Since the intersecting fault planes and different slip directions occur in a relatively small area and in a very short time span, we suggested a complex and segmented fault system for the area, which might imply either a complex stress situation prior to the events or significant lateral variations in shear strength along the fault area.
The second event studied in this thesis was the Kerman triplet of 2017 (Mw 6.1 - 6.0) in eastern Iran. The triplet occurred at the eastern border of a
re-Figure 5.1.3D slip map of the fault planes for (a) the 2012 Varzeghan doublet, (b) the 2017 Kerman triplet.
straining bend, where the western border had been identified as a reverse fault in 2005. Using the Haskell line source directivity definition and a simple de-convolution method, we could detect the rupture propagation direction of the large earthquakes (M ≥ 5) in the sequence. Based on our directivity analy-sis, a predominant rupture propagation direction from southeast to northwest is attributed to the sequence, and we consider that as a result of the regional stress field controlling rupture directions in this area. Focal mechanism solu-tions show predominant reverse faulting mechanisms for all three main events, but several differences have been detected in the rupture processes of the three mainshocks (such as hypocenter depth, slip duration, slip distribution pattern) during our slip inversion analysis. Moreover, using Coulomb failure stress, we assessed the stress interaction between the mainshocks and suggested a model for the faulting configuration (Fig 5.1b) and hence the mechanical evolution of the area. We suggest that the third mainshock formed a barrier which pro-hibited further rupture propagation during the first two mainshocks, but which then ruptured after a short delay because of the stress redistribution. As we see, the 2017 triplet yielded a chance to assess faulting mechanisms and rup-turing processes within a restraining bend, which showed activation of faults in the bend area in order to link the surrounding fault systems together.
The 3D representation of the fault planes for the doublet and triplet main-shocks (Fig 5.1a , b) have common features, notably intersection of the fault planes at depth, where there were small to zero slip during the previous event. In the other words, though intersecting, slip was not repeated on the same fault segment during the successive earthquakes.
Outlook
Multiplets are commonly known from subduction zones, often with larger magnitudes (M ≥ 7), distances between their centroids which may exceed tens of kilometers, and time delays between events as large as a few years. The two multiplet sequences studied here are intracontinental multiplets which are very close in epicentral distance, 3 to 10 km, and occurred in a very short time span, 11 minutes to 11 days. Both the sequences showed a complex interaction between strike-slip and reverse faults which characterizes the active tectonic zones in convergent plate-tectonic settings (Berberian and Yeats, 1999). Due to these complexities the seismic hazard may be difficult to quantify in these areas. Meanwhile, not only classical methods used to estimate earthquake haz-ard (such as concept of a seismic cycle), but also modern earthquake rupture forecast models (such as the UCERF3 model for California, Field et al., 2014), do not include the potential of multiple fault rupture in their analysis, possibly leading to misleading hazard assessment.
We suggest that future studies to update the hazard maps and stress regimes in such complex continental regions need to consider the probability of mul-tiplets and the possibility of intersecting fault planes. They also need to be cautious in assuming rigid crustal blocks, and take into account the transfer of shear strain from the main fault to reactivation of hidden faults, many kilome-ters away. It is also important to note that two successive events, even when very close in time and space, can exhibit noticeably different rupture behav-ior. A simple and practical way suggested here is the Coulomb stress analysis, which in spite of its assumptions about the pre-existing stress state provides a useful tool to describe the stress interaction between the faults.
6. Sammanfattning på svenska
”Skalvmultiplette” är sekvenser av större (magnitud M ≥ 6) jordbävningar som inträffar nära varandra i tid, rum och magnitud (t.ex. Astiz Kanamori, 1984). Skalvmultipletter är inte vanliga men de ökar signifikant den seismiska risken i de områden där de inträffar. Ett viktigt steg i bedömningen av den seis-miska risk som skalvmultipletter utgör är att undersöka deras källegenskaper och de mekanismer som ligger till grund för att de sker. I denna studie under-sökte vi källegenskaper och källmekanismer för två skalvmultipletter i Iran, vilka båda inträffade i oväntade områden men i närheten av större förkast-ningszoner. Den första var en dublett som skedde i nordvästra Iran den 11 augusti 2012 (Mw 6.5 och Mw 6.4) och den andra en triplett i östra Iran den 1:a och 12:e december 2017 (Mw 6.0 - 6.1).
I artikel I studerade vi skalvdubletten i nordvästra Iran 2012 som orsakade mer än 300 dödsfall och en mängd skadade. Det första huvudskalvet, med magnitud Mw 6.5, var av horisontalförskjutande typ och associerades med ett 12 km långt öst-västligt markgenomslag. Det andra huvudskalvet, Mw 6.4, karakteriserades av sned glidning på en reversförkastning men utan markgenom-slag eller efterskalv som definierade förkastningsplanet. Teleseismiskt data från mer än 100 P- och S-vågor användes för att invertera för förskjutnings-fördelningen på det första huvudskalvets förkastningsplan. Både P- och S-vågor användes för att stabilisera inversionen och markgenomslaget gav yt-terligare bivillkor till förskjutningsmodellen. Förskjutningsmodellen gav två väldefinierade områden med rörelse, ett centralt område och ett sidoområde. De största förskjutningarna skedde i det centrala området från nära markytan ner till 10 km djup. I sidoområdet, på 12 – 18 km djup i förkastningens västra del, var förskjutningen mindre och med en annan rörelseriktning än i den cen-trala delen. Möjligen representerar sidoområdet rörelser på sekundärförkast-ningar som skedde i samband med skalvet. Med hjälp av förskjutningsfördel-ningen beräknades Coulombspänningsförändringen för det andra huvudskal-vets nodalplan och för de största efterskalven. Med antagandet att den statiska spänningsförändringen från det första skalvet utlöste det andra visar resultaten av Coulombberäkningarna att det nord-sydliga nodalplanet bör ha varit mer instabilt, och därmed sannolikt det korrekta skalvplanet.
I artikel II studerade vi i vilken riktning brottet utbredde sig på förkast-ningarna i skalvtripletten 2017 i östra Iran. Vågformsdata från regionala avstånd
(90 - 400 km) och avfaltningsmetoden med Empirical Green’s Functions (EGF) användes för att undersöka brottriktningsrörelsen på de tre huvudskalven (ML 6.0 - 6.1) och de sex största efterskalven. Två olika metoder användes för respektive P- och S-vågsdata. P-vågsdata användes till att beräkna relativa källtidsfunktioner och med dessa undersöka riktningsberoende variationer i brottets utsträckning i tid. Samma fenomen undersöktes med S-vågsdata genom att studera hur amplitudspektra varierade med riktningen från förkastningen. Resultaten visar att brottet i de två första skalven startade i sydost och utbredde sig åt nordväst. I det tredje skalvet startade brottet centralt på förkastningen och utbredde sig både åt nordväst och sydost. Riktningsanalysen av efterskal-ven visade att fyra av sex efterskalv också hade brottutbredning från sydost mot nordväst. Jordskalvssekvensen som helhet dominerades alltså av brottut-bredning från sydost till nordväst, vilket tyder på att det regionala spännings-fältet kontrollerar brottutbredningen.
I artikel III undersöktes förkastningsparametrar och förskjutningsfördel-ningen på de tre huvudskalven i tripletten 2017 med hjälp av teleseismiska volymvågor. Fler än 50 vågformer från P- och SH-vågor för varje skalv användes i inversionerna och resultaten anpassades till de skalära seismiska momenten från gcmt-lösningarna samt till observationer (eller bristen på de-samma) av markgenomslag. Fokalmekanismerna visar att reversförkastnings-rörelse dominerar hos alla tre skalven men endast det sista producerade ett markgenomslag. Skillnader i brottprocessen detekterades hos de tre skalven, bland annat hade det andra skalvet en signifikant längre brottutbredningstid än de övriga (12 s istället för 6 - 8 s), det tredje skalvet hade ett grundare hypocentrum (5 - 6 km jämfört med 8 - 9 km) och spänningsfallet var också högre i det tredje skalvet. Förskjutningsfördelningarna i skalven användes för att beräkna Coulombspänningsförändringar på förkastningarna. Denna spänningsväxelverkan mellan huvudskalven tyder på att det sydvästlutande nodalplanet i det första skalvet, och det nordostlutande nodalplanet i det andra skalvet, är de korrekta skalvplanen.
7. Acknowledgements
The first time I visited the Uppsala University was in Google Maps! It captured me. I started picturing myself walking in the streets, riding a bicycle and feeling the breeze on my face. In a year, it became a reality and I cannot explain how cheerful I was. Now here I am, a different me, who thrived on endless struggles and kept hope after all failures for the past nine years. I had awesome people beside me during this wonderful journey and I want to take the chance to thank them.
First, I would like to express my sincere gratitude to my supervisor, Roland, whose wisdom, expertise, generosity, patience, and attitude is something I ad-mire. Whenever I went to him with hopeless results, he brought out something valuable from it and helped me see things in a different light. Special grati-tude goes to Hossein, as I luckily had the chance to be his sgrati-tudent back in my master studies, and also benefited from his endless support and advice during my PhD. Hossein, it was actually your outstanding knowledge in seismology and your professional behavior which motivated me to continue my studies to a PhD. You became a true friend to our family and the one who was al-ways there to help us. I am also thankful to Björn, for his guidance, critical comments, subtle points, and his openness even at busy times.
I would also like to express my deep gratitude to Mohammad and I believe that this success would not have been possible without his generous assis-tance, technical support and practical contribution. And Zoya, thank you so much for granting me the opportunity to be a PhD student at Uppsala Uni-versity, and for a year of intimate supervision, gentle guidance and share of experiences. I lost you as my supervisor, but I gained a valuable friendship. I am particularly grateful to Hemin, for his priceless comments and ideas which have remarkably promoted my research work.
I also appreciate my friends and colleagues at Geocentrum. Special thanks to Rebekka and Ashkan for reviewing this thesis and providing me worthy comments and feedback. To Aggela, with whom we started as office mates and then became friends, thanks for inspiring me when I was frustrated, and for your patience and thoughtfulness. I would also like to thank Claudia, Zeinab, Frederic and Silvia, Karin, Giulia, Sissa, Hamzeh and Ka Lok for the mem-orable moments we had during the meetings, fikas and conferences. Thanks
for Ari for the fun fikas we had during the first year, and thanks to Oli and Christoph for the courses we had together. Many thanks to the geophysics department for making me feel welcomed, including (in no particular order) Azita and Alireza, Sahar and Reza, Saeed and Farzaneh, Faramarz, Maria and Michael, ChunLing, Darina, Daniel, Magnus, Monika, Bojan, Remi, Tegan, Georgiana. Thanks to Peter, Emil and Arnaud who kindly helped me with technical problems; to my friends in Hydrology: Farzad, Saba and Faranak; to Taher and Anita for their kind hospitality. Also thanks to the new generation: Mohsen for your humour which was always refreshing me, Alex and Joshi for welcoming me to your office, Mahshad for your candid friendship, and to Ayse, Alba, Christian, Ruth, Paula, Laura, Sebastian, Ruixue, Michael, Mag-dalena, George, Jan, Tatiana, and those I might miss to mention, for providing a friendly environment.
I am grateful to my friends in Sweden, Marie and Omid, Behnoush and Ashkan, Mahshid and Behzad, Mahsa and Hossein, who offered me their priceless friendship and support and with whom I have shared laughter, frus-tration and companionship.
I also want to thank some people from my undergraduate studies. Dr. Sharghi and Dr. Shadmanaman who introduced me to geophysics in a very attractive way and encouraged me to pursue my studies in seismology. Also special thanks to Najmeh, my friend from the master program with whom we shared a lot of scientific and non-scientific discussions and she provided me with all kinds of data I needed during my PhD studies.
There are no words to express my gratitude to my family, to my Mom and Dad, and my lovely brother, Sahand, for their unconditional love, support, en-couragement and prayers they have sent my way. To Mohsen’s family and especially his Mom, for all her help with the kids when we needed it the most. I deeply appreciate the warm hospitality of my uncle Hossein and Mitra, es-pecially when I arrived in Sweden for the first time and of course our visits all after. My deepest gratitude goes to my life time friend and love Mohsen, who stood by me throughout this long journey, patiently listened to me, supported me, encouraged me and was always there for me at the end of the day. To my lovely angels, Ariana and Romina, you cannot imagine how grateful I am for having you in my life. Your shiny eyes, warm hugs and innocent kisses wipe out all my tiredness and are the most powerful treatment for my hopeless moments. Thank you for making me stronger and better and for inspiring me to thrive.
References
Ambraseys, N.N. and Melville, C.P. (1982). A History of Persian Earthquakes. Cambridge University Press, UK.
Astiz, L. and Kanamori, H. (1984). An earthquake doublet in Ometepec, Guerrero, Mexico. Physics of the Earth and Planetary Interiors, 34(1-2):24–45.
Bassin, C., Laske, G., and Masters, G. (2000). The current limits of resolution for surface wave tomography in North America. EOS Trans AGU, 81:F897.
Berberian, M., Asudeh, I., and Arshadi, S. (1979). Surface rupture and mechanism of the Bob-Tangol (southeastern Iran) earthquake of 19 December 1977. Earth and Planetary Science Letters, 42(3):456–462.
Berberian, M., Jackson, J., Fielding, E., Parsons, B., Priestley, K., Qorashi, M., Talebian, M., Walker, R., Wright, T., and Baker, C. (2001). The 1998 March 14 Fandoqa earthquake (Mw 6.6.) in Kerman province, Southeast Iran: Re-rupture of the 1981 Sirch earthquake fault, triggering of slip on adjacent thrusts and the active tectonics of the Gowk fault zone. Geophysical Journal International,
146(2):371–398.
Berberian, M. and Yeats, R. (1999). Patterns of Historical Earthquake Rupture in the Iranian Plateau. Bulletin of the Seismological Society of America, 89(1):120–139. Calderoni, G., Rovelli, A., Ben-Zion, Y., and Di Giovambattista, R. (2015).
Along-strike rupture directivity of earthquakes of the 2009 L’Aquila, central Italy, seismic sequence. Geophysical Journal International, 203(1):399–415.
Calderoni, G., Rovelli, A., and Singh, S. K. (2013). Stress drop and source scaling of the 2009 April L’Aquila earthquakes. Geophysical Journal International,
192(1):260–274.
Christensen, D. H. and Ruff, L. J. (1985). ANALYSIS OF THE TRADE-OFF BETWEEN HYPOCENTRAL DEPTH AND SOURCE TIME FUNCTION. Bulletin of the Seismological Society of America, 75(6):1637–1656.
Das, S. and Kostrov, B. (1994). Diversity of solutions of the problem of earthquake faulting inversion; application to SH waves for the great 1989 Macquarie Ridge earthquake. Physics of the Earth and Planetary Interiors, 85(3-4):293–318. Djamour, Y., Vernant, P., Nankali, H. R., and Tavakoli, F. (2011). Nw iran-eastern
turkey present-day kinematics: Results from the iranian permanent gps network. Earth and Planetary Science Letters, 307(1):27 – 34.
Donner, S., Ghods, A., Krüger, F., Rößler, D., Landgraf, A., and Ballato, P. (2015). The Ahar-Varzeghan earthquake doublet (Mw 6.4 and 6.2) of 11 August 2012: Regional seismic moment tensors and a seismotectonic interpretation. Bulletin of the Seismological Society of America, 105(2):791–807.
Felzer, K. R. (2004). A Common Origin for Aftershocks, Foreshocks, and Multiplets. Bulletin of the Seismological Society of America, 94(1):88–98. Field, E., Arrowsmith, R., Biasi, G., Bird, P., Dawson, T., Felzer, K., Jackson, D.,
