Article,
Statistical analysis of the induced Basel 2006 earthquake sequence: introducing a probability-based monitoring approach for Enhanced Geothermal Systems
Geophysical Journal International, 186 (2): 793--807 (August 2011)
DOI: 10.1111/j.1365-246X.2011.05068.x
Abstract
Geothermal energy is becoming an important clean energy source, however,
the stimulation of a reservoir for an Enhanced Geothermal System
(EGS) is associated with seismic risk due to induced seismicity.
Seismicity occurring due to the water injection at depth have to
be well recorded and monitored. To mitigate the seismic risk of a
damaging event, an appropriate alarm system needs to be in place
for each individual experiment. In recent experiments, the so-called
traffic-light alarm system, based on public response, local magnitude
and peak ground velocity, was used. We aim to improve the pre-defined
alarm system by introducing a probability-based approach; we retrospectively
model the ongoing seismicity in real time with multiple statistical
forecast models and then translate the forecast to seismic hazard
in terms of probabilities of exceeding a ground motion intensity
level. One class of models accounts for the water injection rate,
the main parameter that can be controlled by the operators during
an experiment. By translating the models into time-varying probabilities
of exceeding various intensity levels, we provide tools which are
well understood by the decision makers and can be used to determine
thresholds non-exceedance during a reservoir stimulation; this, however,
remains an entrepreneurial or political decision of the responsible
project coordinators. We introduce forecast models based on the data
set of an EGS experiment in the city of Basel. Between 2006 December
2 and 8, approximately 11500 m3 of water was injected into a 5-km-deep
well at high pressures. A six-sensor borehole array, was installed
by the company Geothermal Explorers Limited (GEL) at depths between
300 and 2700 m around the well to monitor the induced seismicity.
The network recorded approximately 11200 events during the injection
phase, more than 3500 of which were located. With the traffic-light
system, actions where implemented after an ML 2.7 event, the water
injection was reduced and then stopped after another ML 2.5 event.
A few hours later, an earthquake with ML 3.4, felt within the city,
occurred, which led to bleed-off of the well. A risk study was later
issued with the outcome that the experiment could not be resumed.
We analyse the statistical features of the sequence and show that
the sequence is well modelled with the Omori-Utsu law following the
termination of water injection. Based on this model, the sequence
will last 31+29/-14 years to reach the background level. We introduce
statistical models based on Reasenberg and Jones and Epidemic Type
Aftershock Sequence (ETAS) models, commonly used to model aftershock
sequences. We compare and test different model setups to simulate
the sequences, varying the number of fixed and free parameters. For
one class of the ETAS models, we account for the flow rate at the
injection borehole. We test the models against the observed data
with standard likelihood tests and find the ETAS model accounting
for the on flow rate to perform best. Such a model may in future
serve as a valuable tool for designing probabilistic alarm systems
for EGS experiments.
