| Abstract |
That seismic wave speeds in geothermal systems might change with time is an expected consequence both of natural processes and of industrial activities such as fluid withdrawal and injection, and indeed detections of such changes have been reported, based on repeated MEQ seismic tomography. Seismic tomography suffers, however, from limitations that cast doubt on some of these reports. Specifying a three-dimensional model of seismic wave speed in even a small volume requires at least thousands of parameter values, but because of the limited distribution of both earthquakes and seismometers, real data sets virtually never contain enough information. Seismic rays tend to occur in bundles that sample the Earth very unevenly, and standard inversion methods impose assumptions that result in structural features that mimic these bundles. Any changes in the ray distributions, such as those caused by seismicity changes, can therefore produce artificial apparent changes in the wave speeds. It is possible to overcome this difficulty by simultaneously inverting data sets from different epochs, imposing constraints to minimize temporal and spatial variations. This strategy greatly reduces both complex structure and temporal variations that are not actually required by the data. We report the results of applying program TOMO4D, which uses this strategy, to seismic data from Mt. Etna, in Sicily (where strong temporal variations have been reported in the literature), Long Valley caldera, California, which is largely in a natural state, and the heavily exploited Coso geothermal area, California. For Coso, we studied in detail the epochs 1996-2006 and 2007-2012, determining Vp, Vs and Vp/Vs. In our presentation, we will report out latest results. |