Julian, B.R. and G.R. Foulger, Time-dependent seismic tomography of the Coso geothermal area, 1996-2004, Geothermal Resources Council Annual Meeting, Reno, Nevada, 25th – 28th September, 2005.
Time-dependent Seismic Tomography of the Coso Geothermal Area, 1996-2004
Bruce R. Julian, U. S. Geological Survey, Menlo Park, California
G. R. Foulger, Dept. of Earth Sciences, Univeristy of Durham, Durham, U. K.
Keith Richards-Dinger and Francis Monastero, Geothermal Program Office, US Navy, China Lake, California
Studies at several geothermal areas show that (1) geothermal reservoirs often exhibit strong anomalies in the ratio, Vp/Vs, of the seismic-wave speeds, and (2) economic exploitation can cause measurable changes in the wave speeds, which may provide valuable information for monitoring reservoir processes critical for sustained production. The U. S. Navy’s permanent 18-station network of three-component digital seismometers at the seismically active Coso geothermal area, California, provides high-quality microearthquake (MEQ) data that are well suited to investigating possible temporal variations in structure related to processes within the geothermal reservoir for the period from 1996 through 2004.
We winnowed the large data set of arrival times in the Navy MEQ catalog to eliminate redundant and low-quality data, ranking events in terms of the number and geographic distribution of seismic readings available and the goodness-of-fit to the arrival times for each event. This winnowed data set comprised 4811 earthquakes and 79,822 P- and S-phase travel times. First, we inverted the entire selected data set to estimate the average one-dimensional (wave speeds vs. depth) structure. Second, we used this structure as a starting model and inverted the entire winnowed data set to obtain a time-averaged three-dimensional model. Finally, we inverted the data for each of the years 1996 – 2004, separately, using the averaged three-dimensional model as the starting point in each inversion. We repeated this three-dimensional inversion procedure twice for each year: first with a 2-km horizontal grid spacing and then with a 1-km horizontal grid. In all cases the vertical grid spacing was 1 km.
Summary of Results
- Compared with the crustal model currently used by the US Navy for routine earthquake location, the average one-dimensional crustal structure has somewhat lower wave speeds at shallow depth and significantly higher ones at greater depth. Use of a model with higher wave speeds in general will result in smaller estimated earthquake hypocentral depths.
- The area is characterized by relatively high Vp/Vs in the upper ~ 4 km, beneath which the value of this ratio decreases. This change is most likely caused by petrological variations.
- Low Vp and Vs wave speeds are associated with sediments in the Coso Wash and Coso Basin. A Vp/Vs low occupies the northern and eastern part of the geothermal field near the surface (~ 1 km above sea level), and the northern and southern parts of the field near sea level.
- Independent inversions for each of the years 1996 – 2004 separately show, for the 2-km horizontal grid spacing, an irregular strengthening with time of the negative Vp/Vs anomalies between the surface and approximately sea level. This progressive reduction in Vp/Vs results predominately from a increase of VS with respect to VP. Such a change is expected to result from effects of geothermal operations such as decreasing fluid pressure and the drying of argillaceous minerals such as illite.
- Inverting the data with a 1-km horizontal grid spacing produced noisier results without significantly reducing data variance. From this we conclude that the density of seismic stations is insufficient to improve horizontal spatial structural resolution beyond about 2 km.
last update 3rd February, 2006