| Title | Probabilistic approaches in EGS seismic hazard assessment |
|---|---|
| Authors | Jupe, A; Francis, D; Gehrmann, M |
| Year | 2016 |
| Conference | European Geothermal Congress |
| Keywords | EGS, microseismic, seismicity, hazard |
| Abstract | We discuss the growing interest in probabilistic based approaches for seismic hazard assessment in EGS (Engineered Geothermal Systems), in which they are used both for estimating the likely maximum event size and also as the basis of real-time adaptive “traffic light” systems. Probabilistic approaches are also of practical interest to EGS developers because they do not demand a detailed knowledge of fault parameters, hydro-mechanical interaction within the reservoir or the seismic rupture process. However probabilistic approaches not only require an apriori statistical sample, but also confidence that large events are not somehow “anomalies” or “special” events, but are part of a continuous statistical and scalable seismicity distribution. This argument is made difficult by the frequent distinction between “induced” and “triggered” seismicity. The implication being that the former are part of some continuous statistical distribution, whereas the latter are anomalous seismicity related to some specific, and sometimes unknown, large-scale structure. To illustrate this latter point we then consider a specific example of apparently “anomalous” seismicity from the UK Hot Dry Rock EGS Project (Rosemanowes) in the 1980’s. The largest event was ML 2 and was felt locally. At the time the largest events appeared distinct in terms of both magnitude and focal mechanism compared to the 5000+ other “induced” events. Consequently there was ongoing uncertainty as to whether the larger events were in fact anomalies or a statistically probable consequence of the long-term injection operations. A present-day implication of this uncertainty is whether it is appropriate to adopt probabilistic based approaches for EGS development currently planned in the area, or whether a purely mechanistic approach is required that takes into account local geological and tectonic conditions, with its high associated exploration demands. Recent re-analysis of the data suggests that the Rosemanowes events were in fact consistent with a general statistical relationship between seismic energy release and injected fluid volume, and also with the range of focal mechanisms expected from the interaction of injected fluid, stress regime and fracture distribution. We conclude that the large events observed at Rosemanowes should not therefore be considered anomalous, but as a statistical and geomechanical consequence of overall net fluid loss within the subsurface. This example not only illustrates the care that needs to be taken when distinguishing between induced and triggered seismicity, it also provides some support for the use of probabilistic based hazard assessment approaches and the potential of real-time adaptive “traffic light” systems. |