Stefan Baisch

Induced Seismicity during the Stimulation of a Geothermal hfr Reservoir in the Cooper Basin, Australia

A long-term fluid-injection experiment was performed in the Cooper Basin (Australia) in 2003 to stimulate a geothermal reservoir. More than 20,000 m 3 of water were injected into the granitic crust at 4250 m depth. During reservoir stimulation about 27,000 induced seismic events were detected by a local, eight- station seismic monitoring system deployed in nearby boreholes. Hypocenter locations for 11,068 events were determined by using an averaged velocity model that was calibrated by associating early events with the injection point. The spatial hypocenter distribution forms a nearly subhorizontal structure with a lateral extension of 2 km × 1.5 km and an apparent thickness of approximately 150–200 m, which is in the order of the hypocenter location confidence limits. The hypocenter distribution exhibits a high degree of spatiotemporal ordering with the seismic activity systematically migrating away from the injection well with increasing time. Previously activated regions become seismically quiet indicating relaxation processes. High-resolution relative hypocenter locations determined for clusters of “similar” events locally reduce the apparent thickness of the structure to the level of a few tens of meters indicating that the reservoir is dominated by a single fracture zone only. Consistent with these findings, a subsequently drilled well intersects a dominating, high-permeable fracture within 15 m of the predicted intersection depth. Based on drilling and logging information, the fracture zone is interpreted as a preexisting (possibly tectonically formed) feature that (partly) sheared during stimulation. Triggering of the induced seismicity is found to be predominantly controlled by the increase of fluid pressure implicating a (local) reduction of the effective normal stress resolved on the fracture plane. Additionally, perturbations of the stress field caused by the largest-magnitude events may trigger seismicity (“aftershocks”) on a local, short-ranging scale.

Seismic waveform attributes before and after the Loma Prieta earthquake: Scattering change near the earthquake and temporal recovery

During 1987–1995 several clusters of nearly identical seismic events (multiplets) occurred near the Loma Prieta source region. These multiplets allow us to investigate and demonstrate spatial and temporal changes in seismic wave character associated with the 1989 Loma Prieta main shock. For seismogram pairs we use a moving window technique to compute coherencies depending on lapse time and frequency. Post-Loma Prieta events have reduced coherencies with pre-Loma Prieta events in a spatially limited region close to the Loma Prieta hypocenter, while other paths remain nearly unaffected. These changes gradually recover within a time interval of 5 years after the Loma Prieta earthquake. A possible explanation for the time dependence is coseismically opened cracks which cause scattering increase for wavefields after the Loma Prieta event. Postseismic relaxation processes such as crack healing, fluid diffusion, or after deformations lead to progressive closure of these cracks with time after the main shock. Thus the scattering properties of the local crust approach the pre-main shock state.

Probing the Crust to 9-km Depth: Fluid-Injection Experiments and Induced Seismicity at the KTB Superdeep Drilling Hole, Germany

A 60-day, long-term fluid-injection experiment was performed at the 9.1-km-deep Kontinentale Tiefbohrung, Germany (KTB), borehole. About 4000 m3 of water were injected into the well head to induce seismicity near the open-hole section at 9-km depth. Because of several leaks in the borehole casing (unknown before), seismicity occurred at distinct depth levels between 3-km and 9-km depth. Two events occurred at 10-km and 15-km depth. The combination of a temporary, 40-element, three-component surface network of seismometers and a three-component downhole sonde at 3.8-km depth in the nearby pilot hole enabled us to determine absolute hypocenter locations by using a velocity model that was calibrated from several downhole shots at depths of 5.4 km and 8.5 km. Of a total of 2799 induced events, hypocenter locations were obtained for 237 events having good signal-to-noise ratio at surface stations. The spatiotemporal distribution of hypocenters at each depth level exhibits complex structures extending several hundred meters from the injection points, with strong spatial and temporal clustering. Regions that were seismically active at a certain time often showed reduced or no activity at later times, indicating local shear-stress relaxation. A similar “memory” effect (Kaiser effect) is observed by comparing hypocenter locations of the present experiment with those obtained for a previous injection experiment at KTB. The limitation of hypocentral depths to 9.1 km for events near the borehole suggests changes in rheological properties of the upper crust and thus supports a transition from the regime of brittle failure to ductile deformation at this depth. Large fluid-level changes observed in the nearby pilot hole demonstrate that fluid flow occurs over distances greater than 1.5 km and that major flow zones are not mapped by the induced seismicity. This might also explain the occurence of isolated events at greater distances and depths. Brittle failure at depths greater than 10 km indicates the existence of critically stressed fractures even at temperature over 300°C.

Spectral analysis with incomplete time series: an example from seismology

Abstract A method for spectral analysis of nonequidistantly spaced time series is presented: the CLEAN algorithm performs an iterative deconvolution of the spectral window in the frequency domain. We demonstrate the capability of the method on synthetic data examples and apply CLEAN to seismological data, in an example where we seek temporal changes in elastic wave velocities. The observed periodic changes of phase differences consist of frequencies, which in principle can be explained by the influence of solid earth tides, but also by other effects with similar periodicities. Only CLEAN enabled us to enlarge the time window over missing data segments until the frequency resolution was accurate enough to rule out solid earth tides as cause for the observed periodic changes. A MATLAB version of the CLEAN algorithm is available from the authors, or from the IAMG server.

Earthquake Cluster: What Can We Learn from Waveform Similarity?

Abstract The seismograms of earthquakes, which have closely spaced hypocenters, tend to be similar due to the similarity of the Green’s functions characterizing the source-receiver paths. Based on the λ /4 criterion, it is frequently argued that similar earthquakes may represent repeated slip of the same patch of a fault. Because of the phenomenological nature of waveform similarity, such interpretations are strongly dependent on seismic signal characteristics and on the way, the waveform similarity is defined. In this article, we use two-dimensional synthetic wave-field simulations in lateral heterogeneous media to investigate how the waveform similarity of closely spaced hypocenters changes with interevent separation. We analyze the influence of correlation window length, signal frequency bandwidth, and source-receiver geometry on the waveform similarity and discuss under which conditions the λ /4 criterion can be applied to the synthetic data set. With the correlation window length defined as 2.8 times the travel-time difference between the S - and P -phase onsets, we find a correlation threshold value of 0.95 independent of the signal frequency bandwidth. We use the same threshold value for two field data examples that are similar to the synthetic data in frequency content and waveform complexity, and we discuss the implications of the λ /4 criterion. For three microearthquakes occurring during a fluid-injection experiment at the German deep drilling site (Kontinentale Tiefbohrung [KTB]), the interevent separation constrained by the λ /4 criterion is sufficient to identify these events as a sequence of repeating earthquakes in the sense that at least a fraction of the source area experienced repeated slip. For a second data example of four natural (micro-) earthquakes occurring near the island of Crete, the λ /4 criterion does not sufficiently constrain the hypocenter location to identify these events as repeating earthquakes due to the lack of high-frequency information.