dc.description.abstract
The debate on self-similarity of source physics between small and large earthquakes has
been high on the agenda in earthquake science for decades. Seismic source parameters,
such as corner frequency, seismic moment, and stress drop, to name only a few, have
been analyzed and compared across the entire earthquake magnitude range, considering
large natural events, induced seismicity, and laboratory acoustic emissions. Stress drop
is an important source parameter that substantially affects ground motions and is often
used as an earthquake scaling parameter. The majority of stress drop studies show on
average independence on earthquake size and thus scale-invariance is assumed. Scaling
breakdown has, however, also been reported for individual datasets in which stress
drop scales with earthquake magnitude. Regardless of the debate on self-similarity,
a large global stress drop scatter of 0.01-100 MPa has been observed throughout all
earthquake sizes, which is not yet fully understood. The reasons for this scatter need
to be investigated in more detail to better understand the similarities and differences
between earthquake source characteristics and the corresponding physical processes in
the earthquake source.
This doctoral thesis focuses on small earthquakes, including induced seismicity and
laboratory acoustic emission events, and addresses two groups of factors whose effects
are assumed to cause scatter in stress drop estimates: 1) Non-physical factors that cover
error propagation, and inappropriate assumptions made during the analysis of seismic
data, can lead to bias in source parameter estimates. 2) Physical factors, which reflect
the actual properties of earthquakes foci and rupture processes, are assumed to naturally
influence the resulting static stress drop. In this thesis, non-physical effects on stress drop
estimates are suppressed by considering only high-quality data, adequate methodologies,
and carefully evaluated and selected parameters for analysis. One aspect that this thesis
studies in more detail, is the effect of high-frequency wave attenuation on seismic
records. Attenuation is expressed as the quality factor Q, which describes the quality of
the medium that is in general unknown. Q is important for the correction of amplitude
source spectra while estimating source parameters. Incorrect assumptions of Q can bias
the high-frequency spectral fall-off leading to error propagation (a non-physical effect)
during further data analysis.
The first study analyzes the time-dependent decay of S-coda waves of induced seismicity
at The Geysers geothermal field, California, to better assess attenuation effects for
high-frequency seismic signals. The application of the moving window method enables
us to estimate stable coda quality factors (QC) for pre-defined frequency bands in the
frequency domain. An additional sensitivity analysis is conducted beforehand to evaluate the impact of different parameters used in the coda technique, which are the magnitude
range, signal-to-noise ratio, maximum permitted uncertainties of QC estimates, moving
window width, lapse time effect, and total coda length. Findings of the sensitivity
analysis emphasize the relevance of parameterization and the usage of high quality data.
Waveforms of in total 717 shallow earthquakes with duration magnitudes 1 < MD < 3 of
two spatially separated locations in the northwestern and southeastern part of The Geysers
were investigated. Both areas show clear differences in attenuation properties dependent
on locally varying geological, structural and geothermal characteristics. The coda quality
factor QC was estimated for frequencies up to 70 Hz, exceeding previous field data
considerations. QC obtained from the northwestern dataset was also investigated in the
context of temporal injection variations. The temporal stability of QC suggests that
either the local fault network is on average constant over time or the injection-induced
fractures are not detectable with the frequencies considered. Additionally, QC from the
northwestern dataset is compared to the quality factor obtained by direct S-waves (Qb ).
QC shows on average higher stability in mean estimates compared to Qb . Therefore, it is
suggested that QC leads to fewer uncertainties of further estimated source parameters
such as stress drop.
The second study focuses on acoustic emission (AE) events of two laboratory triaxial
stick-slip experiments on oven-dried Westerly granite samples. Source parameters, in
particular static stress drop, of acoustic emission events from a rough and a smooth
fault were analyzed taking advantage of the spectral ratio method based on a multi-eGf
(empirical Green’s function) approach. Here, attenuation effects (path- and site effects) of
linked, co-located events were suppressed and source parameters were directly obtained
from the source spectra using the quasi-dynamic Madariaga source model. For the first
time, a span of more than three orders of magnitude (-9 < MW < -5.6) was evaluated for
laboratory AE events providing a reasonable base for scaling analysis. Obtained AE
stress drops are comparable to the globally observed estimates. However, a clear scaling
breakdown was observed for both faults putting the global self-similarity assumption
into question. Here, the physical aspects of fault surface roughness, source radius, and
rupture velocity were scrutinized. AE stress drop shows mainly no dependence on source
size and only very little variation with rupture velocity changes. The observed stress
drop–magnitude scaling therefore suggests differences of slip over rough and smooth
fault surfaces. The complexity of rough faults might inhibit larger slips leading to lower
stress drops, whereas smooth faults can result in larger slips and thus higher stress drops
due to fault simplicity.
These two studies provide new insights into the analysis of high-frequency seismic
signals. More comparative studies based on similar approaches are necessary for further earthquake cross-scale investigations. The variety of contemporary analytical methods
and models render the comparison of studies challenging. Therefore, data- and interdisciplinary expertise exchange, data standardization, careful data quality evaluation,
and preservation of methodological applications are needed to facilitate and accelerate
future research. Drawing on these crucial requirements for future investigations, the third
part of this thesis complements the first two scientific studies. It shows the innovative
development of an online research platform headed by the Thematic Core Service Anthropogenic Hazards (TCS AH) within the framework of the Implementation Phase of
the European Plate Observing System (EPOS-IP). The TCS AH team introduces the
genesis of the novel online platform IS-EPOS (IS = Induced-Seismicity) that consists
of a large number of complete, comprehensive datasets and applications related to the
exploration and exploitation of georesources and the underground storage of liquids and
gases. The pioneering work of the EPOS project provides an example of how enhanced
research could deal with the similarities and differences between individual studies, and
thus lead to a more comprehensive understanding of seismic data and its sources.
en
