Content uploaded by Dimitris Dais
Author content
All content in this area was uploaded by Dimitris Dais on Apr 08, 2020
Content may be subject to copyright.
"DIFFERENCES" BETWEEN INDUCED AND NATURAL SEISMIC
EVENTS
İhsan Engin BAL1, Dimitrios DAIS2, Eleni SMYROU3
ABSTRACT
Induced earthquakes exhibit certain characteristics which raise the question whether they are “different” than the
naturally caused seismic events or not. Comparing a natural earthquake event to another already has inherent
difficulties, additionally to the complexities of comparing a natural event to an induced one. This paper presents
the results of an effort to define significant seismic parameters to be used in comparing natural events to induced
ones. Wavelet analyses of various records have been checked in terms of effective period range and duration.
Recorded accelerations have been applied to SDOF URM wall numerical models, calibrated with available
experimental results. Damage history and the accumulation of damage of the analysed walls from induced events
are compared to those found from “similar” natural events. The initial outcome of this effort is that a single
naturally caused seismic event does not provide any meaningful difference as compared to an equivalent induced
one if wavelet content and structural response are concerned. There may be, however, also an effect of
accumulation of damage during repetitive induced events.
Keywords: induced seismicity; comparison of seismic events; masonry structures
1. INTRODUCTION
Induced earthquakes related to large-scale human activities of exploitation of natural resources
constitute one of the most recent research topics in earthquake engineering. Natural resource sites in
Europe and in the US, such as Groningen, Basel, Texas and Oklahoma, attract the research interest
rendering induced seismicity as the spearhead in earthquake engineering at the moment. The main
question arising thus is whether the existing state-of-the-art in earthquake engineering discipline is
applicable to the regions that are introduced to induced seismic events and were not prone to seismic
activities before. This question covers a wide spectrum of topics from seismological definitions and
measurement of earthquakes to response of non-seismically designed structural inventory to recursive
and relatively low magnitude events. The discussion goes back even to the trivial question of whether
these events can be called “earthquakes” or simply are “tremors”.
Comparison of seismic events to each other is at least a difficult endeavour, if possible at all. Earthquakes
are complex geophysical events affected by numerous parameters, most of which are still unknown or
unidentified by the researchers. Apart from the inter-event variabilities observed among seismic events
around the world, there are also large intra-event variabilities observed within the group of seismic
stations triggered by the very same seismic event. An immense effort is spent by researchers to minimize
the inter- but also intra-event variabilities in ground motion prediction models. Considering that even
the comparison of seismic records from two relatively close stations within the same seismic event is
not straight-forward, then comparison of natural events to induced seismicity events which do not occur
1Professor, Hanze Univ. of App. Sciences, NoorderRuimte, Groningen, The Netherlands, i.e.bal@pl.hanze.nl
2PhD Researcher at Hanze UAS, and PhD Candidate at New Castle University, UK, d.dais@pl.hanze.nl
3Assoc. Prof., Hanze UAS, NoorderRuimte, Groningen, The Netherlands, e.smyrou@pl.hanze.nl
2
at similar locations in most cases becomes an extremely challenging task.
One of the major barriers in this comparison is the availability of data. The seismic records from induced
seismic events are, first, derived from a very limited number of regions, and second, they cover a very
short period of observation, which is shorter than a decade in most cases. Furthermore, relatively small-
magnitude earthquakes need to be compared to induced earthquakes, but such low-magnitude seismic
events are not always reported or kept in servers for long time by the data providers in highly seismic
regions.
This paper looks into parameters that are comparable in both types of earthquakes and searches for
“differences” within the very limited database used. Due to the inherent difficulties mentioned
throughout this paper, a comparison is not possible to reach a definitive conclusion of whether natural
and induced seismic events are different or similar. The approach followed in this paper is to detect
trends for natural seismic events and check whether these trends, even with low correlation, are valid
for the induced seismic events or not. In this way, the authors were not able to prove whether the induced
seismic events and natural events are similar, but they were able to show, at least for a very limited
dataset, whether there exist clear “differences”.
1.1 Previous Research
The existing state-of-the-art may not provide clear distinctions between induced and natural seismic
events, however, most of the established rules and approaches of earthquake engineering can serve to
the purposes of research focused on induced seismicity as well. In this respect, it can be said that most
researchers involved in induced seismicity already presume large amount of similarities between the
two types of events.
An interesting study on the topic was published by Whyte and Stojadinovic (2014). They checked
natural events from California and Switzerland in comparison to Enhanced Geothermal System (EGS)
and Injection Well methods that caused earthquakes in Switzerland, California and Texas. They based
their investigation on analyses of spectral values as well as on URM wall analyses based on typical
Swiss construction. Their major finding is that there are no meaningful differences between the natural
and induced seismic events examined.
Due to the nature of the induced seismicity that exhibits resemblance to short duration events, it is worth
mentioning previous research on comparing blast-induced ground motions with natural seismic events.
Research by Dowding (1996), McGarr et al. (1990), Seed and Idriss (1970) show that there are no clear
differences between blast-induced and naturally caused events provided that the magnitudes and
distances are similar.
Gulia (2010) sets criteria to find the contamination in the European earthquake catalogues due to mine
and blast induced activities, employing methods that are mostly based on the use of logic cross-checks
on the catalogue values, such as the time of the events and the frequency of occurrences.
Cesca et al (2013) tailored a method for discriminating the induced seismicity events by using full
moment tensor inversion and decomposition. They established a procedure to analyse a set of natural
and induced events of similar magnitude that occurred in Germany and neighbouring regions. Induced
seismicity is recorded during different mining and/or reservoir exploitations. Moment tensors are
inverted using a multi-step inversion approach. The authors claim that the method was successfully
calibrated.
Dais et al (2017) provides insights on the cumulative damage the events of induced seismicity may cause
on structures. They numerically applied induced seismicity records over simple walls, first individually
and then sequentially, pointing out a level of damage accumulation in the initial small cycles, a potential
difference between natural and induced seismic events. The study was not conclusive, since the existing
numerical models are not calibrated for very small cycles, and the issue needs further research, possibly
3
backed up by experimental evidence.
2. COMPARISON OF THE GROUND MOTION PROPERTIES
2.1 Examining the Strong Ground Motion Data
The natural records used in this paper are extracted from the European Strong Ground Motion Database
(Ambraseys et al., 2002). Constraining several parameters during the record selection yields to very few
or no records. For example, constraining depth, magnitude, soil type and PGA range all at the same time
to values similar to the Groningen records would yield almost no records fitting the criteria. In order to
overcome this problem, the soil type of the station location was fixed to D (CEN, 2004), while the depth
parameter was released and the magnitude parameter was set between 3 and 5. As mentioned earlier, in
the introduction of this paper, the common strong motion databases are not rich in terms of small
magnitude earthquakes.
The induced event records used in this paper are all taken from Groningen gas field. Five records include
the damaging Huizinge earthquake, as recorded in the Middelstum station, as well as records from four
other important events that caused damage to various buildings in the region. The magnitudes of the
selected events vary from 2.6 to 3.6, while the depth is constant, as typically happens in Groningen
earthquakes, at 3km.
One important problem in comparing records to each other is to determine which components to
compare. An acceleration record typically has three components, two of which are horizontal. The
direction of these horizontal components may follow a certain pattern or may be completely random. In
any case, the direction of each component in respect of the epicenter, due to an expected level of polarity,
is also an important factor that cannot be easily controlled when placing the sensors. In brief, it would
be of little interest, if not completely irrelevant, to compare a component of a natural record set to another
one from an induced seismicity in a random sense. A way of normalization, that would minimize the
effects of polarization, is thus required. There are approaches for combining the two components,
intensely used in hazard studies (Baker and Cornell, 2006; Boore et al., 2006), but they are not suitable
for the purposes of this paper since they combine not the recorded motions but the component spectra.
An approach, previously used by Smyrou et al (2011) was used here for normalizing the polarity.
According to that, the horizontal components are rotated 1 degree each time, 180 times in total, and the
strongest (i.e. with the highest PGA) component is found. This new component is used for the
comparisons.
Table 1. List of the natural seismic records used in comparisons.
#
Event
M
Station
R(2)
(km)
EC8 Soil
Class
Depth
(km)
PGA(3)
(m/s2)
1
1976, Friuli, Italy(1)
MW=4.1
BUI
7
D
10
0.39
2
1997, Umbria, Italy
ML=4.2
CLF
7
D
10
0.44
3
1997, Umbria-Marche, Italy(1)
MW=4.3
CLF
2
D
2
0.2
4
1992, Levkas Island, Greece
ML=3.5
LEF
5
D
62
0.24
5
1993, Pyrgos, Greece
MW=4.1
PYR1
6
D
7
0.47
6
1993, Pyrgos, Greece(1)
MW=4.7
PYR1
3
D
10
1.16
7
1988, Pyrgos, Greece
MW=4.2
PYR1
7
D
27
0.43
8
1988, Trilofon, Greece
MW=4.8
THE4
7
D
20
0.36
9
1982, NE Banja Luka, Bosnia&Her.
MS=3.3
BAN2
9
E
10
0.21
10
1980, Skopje, FYROM
ML=4.5
SKOO
22
D
20
1.31
11
1980, Skopje, FYROM
ML=4.5
SKP
23
D
20
0.21
12
1978, Stolac, Bosnia&Herzegovina
MS=3.6
STO
15
E
15
1.04
13
1980, Zergan, Albania
MS=4.3
DEB
15
E
12
0.75
14
1996, Pyrgos, Greece
MW=4.7
PYR1
2
D
0
0.44
(1) Aftershock record (2) Epicentral distance (3) PGA of the rotated component
4
The list of natural events used in this paper are given in Table 1, while a similar list for the induced
events can be found in Table 2. The first step of comparison is based on visual inspection of the recorded
motions. A visual inspection on the acceleration waveforms can provide valuable insights, mostly
qualitative, to experienced eyes. Among the information that can be retrieved are the high- or low-
frequency content related to soft or firm soil conditions, duration, number of significant cycles, or even
existence of liquefaction. The waveforms of the natural and induced seismicity records can be found in
Figure 1and Figure 2, respectively.
Table 2. List of the induced Groningen seismic records used in comparisons.
#
Event
M
Station
R(1)
(km)
EC8 Soil
Class
Depth
(km)
PGA(2)
(m/s2)
1
2012, Huizinge
MW=3.6
MID1
2
D
3
0.88
2
2013, Het Zandt
ML=3.0
BLOP
3.5
D
3
0.68
3
2014, Zandeweer
ML=2.9
BMD2
3.1
D
3
0.76
4
2013, Zeerijp
ML=2.8
BWSE
4.1
D
3
0.25
5
2017, Slochteren
ML=2.6
G460
1.8
D
3
0.35
(1) Epicentral distance (2) PGA of the rotated component
Figure 1. Waveforms of the European earthquake records from natural events used in the comparisons
5
Figure 2. Waveforms of the Groningen earthquake records from induced events used in the comparisons
What is striking at a first glance is that the natural records #6, 10, 11 and 12 in Figure 1 have much
higher acceleration values and significant cycles as compared to the rest of the records, in both groups.
The rest of the natural records in Figure 1, however, do not present any meaningfully difference. Another
point worth mentioning is that record #1 in Figure 2, which is the strong 2012 Huizinge record, provides
more number of cycles with significant acceleration values, and interestingly, with larger dominant
periods. This can be visually detected, but it is later confirmed in this paper with wavelet transformation
plots as well. The Huizinge record, the most damaging event in Groningen (Dost and Kraaijpoel, 2013),
clearly differentiates from the other Groningen records even in the most basic first visual inspection.
The difference of Huizinge event than the other events recorded in the area, in any case, needs further
explanation.
The duration of a strong ground motion record is another important ground motion parameter due to the
short total time and impulsive nature of the small magnitude shallow earthquakes in general. One of the
parameters scrutinized in GMPE models, for example, v4 (Bommer at al., 2017a) and v5 (Bommer at
al., 2017b), specifically devised for the Groningen gas field, is the duration. The significant duration
definition (Trifunac and Brady, 1975; Bommer et al., 2015) was used in GMPE v4 and v5, correlating
magnitude and distance parameters with the duration. Small amplitudes of the induced seismicity do not
mobilize the soil nonlinearity in most cases, causing thus high-frequency pulses on structures. In other
words, if the induced motion is strong enough to trigger high nonlinearity in soil layers resulting in a
few significant high-energy pulses, then the damage potential is deemed to be higher. This reasoning,
however, is no different in natural seismic events. The main difference is that the regions with natural
seismicity are expected to experience larger earthquakes, and the building inventory is thus designed or
prepared to withstand larger strong motion amplitudes than the ones that are comparable with the
induced seismicity events. The duration of the induced seismicity, and the number of non-trivial cycles,
thus becomes an important threshold that can move an induced event from “tremor” to a real
“earthquake”. But again, this is valid for a natural event too.
What has been done in this study is different, namely the duration is not studied over the strong ground
motion records but instead over their effects on a simple wall structure. This is because the duration
definitions in the strong ground motion domain are based on structure-independent parameters, such as
the time range a certain energy fraction is released (Trifunac and Brady, 1975) or a range in which a
threshold acceleration is exceeded the first and then the last time (Kawashima and Aizawa, 1989).
2.2 Wavelet Analyses Results and Observations
Two acceleration records can be compared to each other by using dozens of different properties. This is
why a comparison based on a few parameters, importance of which mostly depend on subjective expert
judgement, is not enough to reach clear conclusions. However, an effort has been made here for
comparing induced seismicity and natural event records based on some of the available strong ground
motion parameters. Wavelet transformation, a strong tool that can present the energy content of a record
in time and frequency domains at the same time, is mainly used.
6
The wavelet analyses presented here have been conducted using Matlab software and employing the
Morlet wavelet (Goupillaud and Morlet, 1984). From the several wavelets in the literature, the Morlet
wavelet is chosen as the most compatible with earthquake motions (Smyrou et al., 2016). It contains
five cycles, with a shape quite similar to a typical ground motion record. The maximum period of interest
in the results depicted in the scalograms is limited to 1sec period. In cases of larger earthquakes, this
can go up to 4sec (Smyrou et al., 2016). In the graphs given in Figure 3 and Figure 4, the abscissa
represents the real time, and the colours convey the match of the record at the relevant time slice with
the pre-defined wavelet, or else the energy content, which is proportional to the level of amplification
in terms of spectral acceleration. The energy content is produced as the multiplication of the amplitude
and the frequency of the wavelet pieces summed up at each time step. The scalograms thus depict the
fraction of the overall energy content (i.e. acceleration multiplied with time) of the record both in the
time and frequency (period) domain. Due to space limitations, only the wavelets from the first 10 of the
natural records listed in Table 1 could be presented here, in Figure 4. The wavelet plots from the induced
seismicity records are given in Figure 3. Please note that only a 6-sec (+-3sec) window is presented here
for the sake of simplicity, where the time zero corresponds to the instant at which the PGA occurs at
each recorded motion.
Record #1 Record #2
Record #3 Record #4
Record #5
Figure 3. Wavelet plots for the records taken from Groningen induced earthquakes, where the rotated strongest
component is used (see Table 2 for the records and events)
7
Record #1 Record #2
Record #3 Record #4
Record #5 Record #6
Record #7 Record #8
Record #9 Record #10
Figure 4. Wavelet plots for the 10 of the records taken from European natural earthquakes, where the rotated
strongest component is used (see Table 1 for the records and events)
One conclusion from Figure 4 is that, despite the fact that they are all natural events and have relatively
close properties, there is no clear pattern that can be detected in the wavelets of these natural events
apart from the record #10. Some of the records (i.e. records #3 to 8) present a tiny energy release in
short period ranges, typically below 0.1sec, and in a very short time ranging from 0.2 to 1 sec. Records
#1 and 2 have similar patterns, while record #10 exhibits larger energy release in longer periods. As
mentioned above, the Huizinge record, as shown in Figure 3, presents different characteristics than the
rest of the Groningen records.
8
Figure 5. Correlation of the dominant periods from wavelet analyses with the magnitude and distance parameters
of the seismic events
In order to better compare the results of the wavelet analyses, a plot, shown in Figure 5, is produced.
The x axis in this plot is the dominant period taken from the wavelet plots, showing the period value at
which the highest energy release was observed. The y axis is a parameter that combines magnitude and
distance values in the same fashion with the old attenuation relationships (Campbell, 1985). According
to that, the y axis consists of magnitude of each event divided by the square of the epicentral distance to
the recording station. It should be noted that the reason such a plot is produced is not to seek for a
calibration but is rather to find out if the induced seismicity records would present any meaningfully
different clustering in this weakly correlated tendency plot. As it can be seen in Figure 5, however, no
clear difference was detected between the records of the induced and natural events examined.
4. COMPARISON OF THE RESPONSE OF A WALL STRUCTURE
4.1 Using Structural Analyses
A simple wall, tested in-plane at EUCENTRE, Pavia, and sponsored by NAM (Nederlandse Aardolie
Maatschappij) is used here for comparisons. The specimen is a full-scale wall, resembling the typical
calcium-silicate unreinforced masonry walls that can be found in the Groningen region. The dimensions
of the wall specimen can be found in Figure 6. More detailed information regarding these tests can be
found in Graziotti et al., 2015 and 2016.
At this point, a simple numerical model was created in Seismostruct software (Seismosoft, 2017) to
simulate the response of the tested specimens for the very first cycles of the conducted experiments,
which seem to be of major importance for the case of induced seismicity. The Ramberg – Osgood
(Ramberg and Osgood, 1943) constitutive model was applied on a link element, taking into
consideration the physical characteristics of the tested specimens, such as the mass, stiffness, and
strength. The mass used in dynamic analyses was calculated by translating the axial load and the self-
weight into mass (Figure 6).
As it can be inferred from the structural characteristics derived from the cyclic tests, the strength of the
wall is considered to be adequate to withstand the frequent but mostly weak seismic events that occur
in the Groningen region (Dais, 2017). As a matter of fact, this led to the necessity to further investigate
the response of the tested specimen not in the range of its ultimate strength but under numerous cycles
of low range of loading, which may finally play an important role in the overall response during a
relatively stronger event (Dais et al., 2017). Specifically, a thorough investigation of the test results
9
pointed out that nonlinear phenomena take place within a range of small top displacements, much before
structural yield displacement. In Figure 6, the first cycles of the conducted tests are presented for the
specimen. It can be noticed that the unloading branches do not follow the loading one, leading thus to a
small amount of residual displacements even for quite low loading. This type of response leads to the
assumption that the frequent earthquakes that strike the structures may produce a noticeable
accumulation of displacements that has the potential to be more destructive when the relatively strong
ground motion occurs, an important aspect of induced seismic events that needs further investigation.
Comparison of the strong motion parameters and wavelet plots can provide insights; however, the
ultimate goal is to find out if the two types of earthquakes would cause different levels of damage to the
same types of structures. This is also a difficult comparison since this time complex structural properties
can intermingle with the strong ground motion parameters. In case structures were analysed, triggering
of different modes by different earthquake records, irregularities, out-of-plane modes in URM, and
several other issues could lead to even higher scatter and lower correlation in any comparative effort. In
order to minimize such effects, a simple wall model was used in this study.
4.2 Comparison of the Cumulative Hysteretic Energy Plots
The calibrated wall model was subjected to natural and induced seismic events studied in this paper.
The base shear-top displacements loops were obtained, and the cumulative hysteretic energy was
calculated for each record. The cumulative energy plot is produced by calculating the area inside the
hysteresis loop from the beginning of the analyses until the moment of interest. The cumulative
hysteretic energy plots, however, cannot be comparable unless normalized both in terms of time and
energy. In order to achieve that, again a 6sec time window is used, as it can be seen in Figure 7. In this
case, time zero corresponds to the instant at which the 50% of the total cumulative hysteretic energy is
attained. The cumulative hysteretic energy values in the y axis are also normalized to be comparable.
This normalization is made by dividing the cumulative energy from the results of each record to the total
hysteretic energy obtained from the first 3x3 cycles shown in Figure 6. This approach allows the
cumulative hysteretic energy imposed by a record to be compared to a nominal structural hysteretic
energy value.
Figure 6. Dimensions of the wall specimen analysed (left), and the calibration of the computer model with the
initial small cycles of the test (right)
The purpose of comparing the cumulative energy plots is to detect how damaging each set of records is,
and in what time range. It can be seen in Figure 7 (left), the accumulation of hysteretic energy occurs in
a range of 20 to 110% of the nominal value used in normalization. This finding alone obviously does
not have any useful meaning. However, one important finding, that is relevant to the purpose of this
paper, is that the cumulative energy plots of induced seismicity records do not have any distinguishable
difference than their natural counterparts, as shown in Figure 7 (left).
One observation over Figure 7 is that the induced seismicity records build up the cumulative hysteretic
energy in a relatively short time. In order to examine this, an effective duration of each record, different
Dimitrios Dais, Ihsan E. Bal, and Eleni Smyrou
3 AVAILABLE EXPERIMENTS ON GRONINGEN-TYPE URM
Experiments sponsored by NAM (Nederlandse Aardolie Maatschappij) and conducted by
EUCENTRE (Pavia, Italy) were carried out on full-scale wall specimens resembling the typical
calcium-silicate unreinforced masonry walls that can be found in the Groningen region. Specif-
ically, two wall specimens representing a slender and a squat wall, correspondingly, were in-
vestigated under in-plane cyclic shear-compression tests. Their characteristic and dimensions
are given in the Table 1, and their experimental configurations are shown in the Figure 4. These
two specimens were chosen to account for different types of wall that can be traced in real
structures are characterized by flexure or shear dominated type of failure.
Specimen L (m) t (m) h (m) σv
(Mpa)
Boundary
Conditions
Slender
Wall 1.1 0.102 2.75 0.52 Double
fixed
Squat Wall 4 0.102 2.75 0.3 Cantilever
Table 1: Characteristics and dimensions of the Tested Specimens.
Figure 5: Experimental configurations considered, Slender Wall (left) and Squat Wall (right) (Pavese et al, 2015
[4]).
The test results presented herein are in terms of Horizontal Force – Drift/Horizontal Dis-
placement (Figure 6 and Figure 7). More detailed results can be found in the aforementioned
report (Pavese et al., 2015, [4]). As it can be inferred from their pseudo-dynamic characteristics
derived from the cyclic tests, their strength should be considered adequate to withstand the
frequent but mostly weak seismic events that occur in the Groningen region. As a matter of fact,
this led to the necessity to further investigate the response of the tested specimens not in the
range of their ultimate strength but in numerous cycles of low range of loading which may
finally play an important role in the overall response during a relatively stronger event. Specif-
ically, a thorough investigation of the tests results pointed out that nonlinear phenomena take
place for very limited value of top displacements. In the Figure 6 and Figure 7, the first cycles
of the conducted tests are presented for the squat wall – the results for the slender wall are
equivalent and thus are not cited in this paper. It can be noticed that the unloading branches do
not follow the loading one, leading thus to a small amount of residual displacements even for
quite low loading. This type of response leads to an assumption that the frequent quakes that
Dimitrios Dais, Ihsan E. Bal, and Eleni Smyrou
2) One year of all records that were recorded at the Middelstum region for in 2014 event
(PGA=0.07g) followed by the Huizinge record of 2012
3) Five times the one-year of bunch that were recorded at the Middelstum region for in
2014 followed by the Huizinge record of 2012
4) Ten times the one-year of bunch that were recorded at the Middelstum region for in 2014
followed by the Huizinge record of 2012
The aforementioned scenarios were chosen this way in order to investigate the impact of the
frequent but of low intensity qu akes that strike the Groningen region in an annual basis to the
stability of the structures when the strong event will follow.
Figure 8: Calibration of the numerical model with the response of the tested specimen in terms of b ase shear ver-
sus top displacement for the first 3x3 cycles of the conducted experiment.
In order to explain the findings, only the results of the scenario #2 is presented here (see
Figure 9 and Figure 10).
Figure 9: Top displacement time history of the squat specimen for scenario #2
10
from the widely known definition by Bommer and Martinez-Pereira (1999), has also been demonstrated
in Figure 7 (right). According to that, the time required to reach from 5% to 95% of the cumulative
energy from each structural response analysis is measured and plotted. These duration values are plotted
against the spectral acceleration of each record, not with the purpose of checking if there is any
correlation between the spectral acceleration demand and the effective duration but better to see if the
results of the induced seismicity records, in any way, differentiate from their natural counterparts. As it
can be seen in Figure 7, no clear difference or separating clustering of data was obtained.
The damage is caused by the induced events on the wall specimen in 1sec on average, while it takes 1.5
seconds to build up the same damage on the same wall by the natural events. This difference, however,
is not found particularly meaningful by the authors because, first, the difference is only a small instant,
and second, the number of induced records is very low to reach a general conclusion about the duration
effect.
Figure 7. Normalized cumulative energy plots of the analyzed masonry wall (left) and the effective durations
within which the 5% to 95% of the cumulative hysteretic energy of the wall response is achieved (right)
5. CONCLUSIONS
In this paper, some of the parameters that are comparable in induced and natural seismic events have
been investigated. “Differences” within the very limited database was examined. Due to the inherent
difficulties mentioned throughout the paper, a firm conclusion is not possible to be reached whether
natural and induced seismic events are different or similar. The approach followed in this paper is to
detect trends for natural seismic events and check whether these trends, even with low correlation, are
valid for the induced seismic events or not. In this way, the authors were not able to prove whether the
induced seismic events and natural events are similar, but they were able to show, at least for a very
limited dataset, that there are no clear “differences”.
In comparisons, it was found that the Middelstum record of the 2012 Huizinge earthquake exhibits
different characteristics than the rest of the Groningen records used in analyses. It was observed that the
cycles of the Huizinge-Middelstum record presented longer period and higher amplitude, also exhibiting
higher number of non-trivial cycles. This effect was clearly shown in the wavelet transformation. The
Huizinge event had the highest magnitude (Mw 3.6) among the five Groningen events investigated. Such
small amplitudes do not mobilize the soil nonlinearity in most cases, causing thus high-frequency pulses
on structures, but the Huizinge-Middelstum record presented a different picture. It needs further
explanation why the Huizinge event presents clear differences as compared to the other Groningen
events.
Strong motion waveforms were compared to each other, and no distinctive pattern was captured.
11
Additionally, wavelet transformation of each record was made, plotting the scalograms, and presenting
the energy release of each record in time and frequency domain. Natural and induced seismicity events
were found not to present clear “differences”.
The selected records were applied on a simple wall structure model, calibrated by using laboratory
experiments. SDOF analysis results were checked in terms of cumulative hysteretic energy plots. These
plots are able to show, in time domain, the development of damage. Building up of cumulative hysteretic
energy, thus the damage, on the same wall during natural and induced seismic events was not different.
It was also checked the time the significant part of the hysteretic energy builds up. It was found that the
difference between the induced seismicity and natural seismicity events in that respect is not significant.
The induced seismicity events are often compared with natural seismicity events, and a quick conclusion
is drawn stating that the induced seismicity events are “different”. This statement is mostly the result of
a misconception of comparing very large seismic events and their results with induced seismicity and
its respective effects. In this case the comparison is made by using non-comparable quantities. This
study is an effort to compare the comparable, starting from the question how to define what quantities
are comparable. A study on a limited database, by using relatively simple comparisons both on strong
motion and structural response fields, yields to the conclusion that there are no clear differences between
induced seismic events and comparable natural seismic events, unless the opposite is proven.
6. ACKNOWLEDGMENTS
Assoc. Prof. Rui Pinho and Asst. Prof. Francesco Graziotti are thankfully acknowledged for helping to
access the experimental data.
7. REFERENCES
Ambraseys N, Smit P, Sigbjornsson R, Suhadolc P and Margaris, B. (2002). Internet-site for European Strong-
Motion Data, European Commission, Research-Directorate General, Environment and Climate Programme.
Baker JW and Cornell CA (2006). Correlation of response spectral values for multi-component ground motions.
Bulletin of the Seismological Society of America, 96 (1):215-227.
Bommer JJ, Martinez-Pereira A (1999). The effective duration of earthquake strong motion. Journal of Earthquake
Engineering, 3(2):127-172.
Bommer JJ, Dost B, Edwards B, Kruiver PP, Meijers P, Ntinalexis M, Polidoro B, Rodriguez-Marek A and
Stafford PJ (2015). Development of Version 2 GMPEs for response spectral accelerations and significant durations
from induced earthquakes in the Groningen field, NAM.
Bommer JJ, Edwards B, Kruiver PP, Rodriguez-Marek A, Stafford PJ, Dost B, Ntinalexis M, Ruigrok E and
Spetzler J (2017b). V5 Ground-Motion Model (GMM) for the Groningen Field, Eds. Jan van Elk & Dirk Doornhof,
NAM.
Bommer JJ, Dost B, Edwards B, Kruiver PP, Meijers P, Ntinalexis M, Rodriguez-Marek A, Ruigrok E, Spetzler J
and Stafford P (2017a). V4 Ground-motion Model (GMM) for Response Spectral Accelerations, Peak Ground
Velocity and Significant Duration in the Groningen field, Eds. Jan van Elk & Dirk Doornhof, NAM.
Boore DM, Watson-Lamprey J, Abrahamson NA (2006). Orientation-independent measures of ground
motion. Bulletin of the Seismological Society of America, 96 (4A): 1502–1511.
Campbell KW (1985). Strong motion attenuation relations: A ten-year perspective. Earthquake Spectra, 1(4): 759-
803.
CEN (2004). Eurocode 8: Design of structures for earthquake resistance. Part 1: General rules, seismic actions and
rules for buildings.
Cesca S, Rohr A and Dahm TJ (2013). Discrimination of induced seismicity by full moment tensor inversion and
decomposition. Journal of Seismology, 17(1):147-163.
12
Dais D (2017). The response of unreinforced masonry to low-amplitude recursive loads: case of Groningen gas
field, MSc Thesis, Istanbul Technical University, Earthquake Engineering and Disaster Management Institute.
Dais D, Bal IE and Smyrou E (2017). Discussion on the response of unreinforced masonry to low-amplitude
recursive loads: case of Groningen gas field. Proceedings of the International Conference on Computational
Methods in Structural Dynamics and Earthquake Engineering, Rhodes, Greece.
Dost B and Kraaijpoel D (2013). The August 16, 2012 Earthquake near Huizinge (Groningen). KNMI Scientific
Report. Royal Netherlands Meteorological Institute (KNMI), Utrecht, The Netherlands.
Dowding CH (1996). Construction Vibrations, Prentice-Hall, Inc., New Jersey.
Goupillaud P, Grossman A, Morlet J (1984). Cycle-octave and related transforms in seismic signal analysis.
Geoexploration, 23:85–102.
Graziotti F, Tomassetti U, Rossi A, Kallioras S, Mandirola M, Cenja E, Penna A, Magenes G (2015). Experimental
campaign on cavity-wall systems representative of the Groningen building stock. Report, EUC318/2015U,
EUCENTRE, Pavia, Italy.
Graziotti F, Rossi A, Mandirola M, Penna A, Magenes G (2016). Experimental characterisation of calcium-silicate
brick masonry for seismic assessment. Brick and Block Masonry: Trends, Innovations and Challenges -
Proceedings of the 16th International Brick and Block Masonry Conference, IBMAC 2016, pp. 1619-1628.
Gulia L (2010). Detection of quarry and mine blast contamination in European regional catalogues. Natural
Hazards, 53 (2): 229-249.
Kawashima K and Aizawa K (1989). Bracketed and normalized durations of earthquake ground acceleration.
Earthquake Engineering Structural Dynamics, 18: 1041–1051.
McGarr A, Bicknell J, Churcher J, Spottiswoode S (1990). Comparison of ground motion from tremors and
explosions in deep gold mines. Journal of Geophysical Research, 95(B13): 21,777-21,792.
Ramberg W, Osgood WR (1943). Description of stress-strain curves by three parameters. National Advisory
Committee on Aeronautics, Technical Note 902.
Seed HB and Idriss IM (1970). Analyses of ground motions at Union Bay, Seattle during earthquakes and distant
nuclear blasts. Bulletin of the Seismological Society of America, 60(1):125-136.
Seismosoft (2017). SeismoStruct - A computer program for static and dynamic nonlinear analysis of framed
structures". Available from URL: www.seismosoft.com.
Smyrou E, Tasiopoulou P, Bal IE and Gazetas G (2011). Ground motions versus geotechnical and structural
damage in the Christchurch February 2011 Earthquake. Seismological Research Letters, 82(6): 882-892.
Smyrou E, Bal İE, Tasiopoulou P and Gazetas G (2016). Wavelet analysis for relating soil amplification and
liquefaction effects with seismic performance of precast structures. Earthquake Engineering Structural Dynamics,
45: 1169–1183.
Trifunac MD, Brady AG (1975). A study on the duration of strong earthquake ground motion. Bulletin of the
Seismological Society of America, 65 (3): 581–626.
Whyte C and B Stojadinovic (2014). Spectral content of induced vs. natural seismicity. Proceedings of the 2nd
European Conference on Earthquake Engineering and Seismology, 24-29 August, Istanbul, Turkey.
