Joint Hypocenter-Velocity Inversion of Local Earthquake Arrival Time Data in Two Geothermal Regions

Lisa Block

Submitted to the Department of Earth, Atmospheric, and Planetary Sciences on April 3, 1991 in partial fulfillment of the requirements for the degree of Doctor of Science

Abstract

P wave and S wave arrival time data from local earthquakes in two geothermal regions, the Los Alamos Hot Dry Rock Reservoir in New Mexico and the Larderello Geothermal Field in Italy, are inverted to simultaneously determine the hypocenter parameters and the three-dimensional velocity structures. In both areas, geothermal production is associated with fractured zones in crystalline rock. The joint hypocenter-velocity inversion yields three-dimensional images of these fractured zones. Constraints are applied which not only stabilize the inversions, but also help the inversions to converge to the most geologically reasonable velocity models. These constraints include spatial velocity derivative regularization, minimum and maximum velocity bounds, and constraints on the velocity values at specified points using information from vertical seismic profiling (VSP) and seismic reflection data.

The Los Alamos Hot Dry Rock Reservoir, located in north-central New Mexico, was created by hydrofracturing in hot crystalline basement rock. The joint hypocenter-velocity inversion is applied to microearthquake arrival time data from nearly 700 events recorded at 4 borehole seismometers during an eight-hour interval of the hydrofracturing. The dimensions of the fractured reservoir, which is the region imaged during the inversion, are less than 1 km. The P wave and S wave velocities are determined independently, except for the indirect coupling through the hypocenters.

The inversion yields low S wave velocity anomalies which generally correlate well with the earthquake locations. The percent velocity perturbations increase as the velocity regularization weighting decreases. Studies of inversions performed using different regularization weightings suggest that the S wave velocities decrease by at least 13% in the most intensively fractured regions of the reservoir. Since the travel time perturbations caused by the fluid-filled fractures are much smaller for P waves than for S waves, the P wave velocities are less constrained by the data than the S wave velocities. For this reason, the P wave velocities are strongly influenced by the velocity regularization and the final VP model is very smooth. Also, because of the limited azimuthal ray coverage, the earthquake locations can trade off with the VP / VS ratios. If the joint inversion is performed without any constraint on the VP / VS structure, then the VP / VS ratios computed from the final velocity models decrease to unreasonably low values (1.1 - 1.3). The earthquake locations are systematically biased by these poor VP / VS values. Inversions performed with a lower bound of 1.60 applied to the VP / VS ratios yield models which satisfy the data as well as the models from the inversions performed without the bound, and they yield more geologically reasonable VP / VS structures. The residuals decrease 11 - 15%. The average absolute change in the earthquake locations during the inversions with the VP / VS bound is 20 - 27 m. The relative earthquake locations are improved by all of the joint inversions performed, even those inversions in which the absolute earthquake locations are biased by poor VP / VS ratios.

The second data set is from the Larderello Geothermal Field, located in west-central Italy. The joint inversion is applied to P wave and S wave arrival times from 269 earthquakes recorded between 1977 and 1990. Most of the earthquakes occur shallower than 8 km depth. The horizontal dimensions of the imaged region are 40 km by 30 km, and the velocities are determined to about 20 km depth. The complex, shallow velocity structure, to 1.5 km depth, is determined from VSP and seismic reflection data. These velocities remain fixed during the joint inversion of the earthquake data. Because there are not enough S wave data to yield an independent, detailed S wave velocity model, the S wave velocities are not directly determined. Rather, the VP / VS ratios are found. Since the S wave velocity structure is not well-resolved by the data, the S wave velocities are required to conform to the P wave velocities. This constraint gives a geologically reasonable "default" model.

The inversion results show three distinct low P wave velocity anomalies within the basement. Two of the anomalies are 5 to 10 km in width and occur between 4 and 7 km depth. The P wave velocities decrease approximately 17 - 25% in these two regions. These features correlate well with the two structural peaks of a strong seismic reflector known as the K horizon. The cause of the low velocities may be hot fluids and/or steam trapped in fractured rock beneath an impermeable zone, as suggested by observations from one well which penetrated the K horizon. This proposed impermeable zone corresponding to the K horizon may be caused by mylonitic deformation along a low-angle normal detachment fault. The third low P wave velocity anomaly occurs at 8 km depth and deeper, and represents the intrusion which is the heat source for the geothermal field. The computed P wave velocities within this feature are as low as 4 - 4.5 km/s, but these values are not well-constrained due to the poor ray coverage at these depths. The location and general shape of the anomaly are consistent with observed gravity data and with an independent velocity model computed from teleseismic arrival time data. The VP / VS model obtained from the joint inversion has poor spatial resolution, due to the small number of available S wave arrival times. For this reason, no distinct VP / VS anomalies are associated with the two small, shallow VP anomalies. An increase in the VP / VS ratio is associated with the large, deep P wave velocity anomaly. This increase in VP / VS, and the corresponding decrease in VP, is consistent with partial melting.