Maximizing EGS earthquake location accuracies

Abstract

Uncertainties in local models of seismic wave speed cause two kinds of errors in estimated microearthquake locations. Errors that differ between nearby events, and which degrade the resolution of faults and similar structures, can now be greatly reduced by relative hypocenter-location methods. Errors in the absolute locations of events, on the other hand, can be reduced using three-dimensional tomographic models of the wave speeds. Uncertainties in local models of seismic wave speed cause two kinds of errors in estimated microearthquake locations. Errors that differ between nearby events, and which degrade the resolution of faults and similar structures, can now be greatly reduced by relative hypocenter-location methods. Errors in the absolute locations of events, on the other hand, can be reduced using three-dimensional tomographic models of the wave speeds, but wave-propagation effects inherently limit the accuracy of tomographic models and therefore the effectiveness of this tactic in contexts such as EGS projects. In typical EGS experiments, fluid is injected one to three km below the surface and monitored by seismometer networks a few kilometers in diameter. Errors in crustal structure produce absolute hypocenter-location errors of the order of a few hundred m. Such errors may preclude use of earthquake locations as targets for production wells. For that purpose, hypocentral accuracies of the order of a few meters are desirable. Adequate accuracies cannot normally be achieved simply by improving the crustal model. A better approach is to measure directly the travel times between the monitoring stations and the stimulation zone, by either (a) firing a shot in an EGS well at the stimulation depth, or (b) deploying a seismometer in an EGS well at the stimulation depth, and firing a shot at each monitoring station. To our knowledge, no such an experiment has yet been performed. Approach (a), firing a shot down-well, risks damaging the well, and furthermore requires explosives that are stable at high temperature. Approach (b) requires sensors and cables that can withstand high temperatures. Nevertheless, the heat tolerance of sensors is improving, and such experiments may now be possible in some cases. We model the likely improvements in earthquake locations that these techniques could provide. We simulate a cluster of earthquakes induced by a theoretical EGS injection, and generate synthetic arrival times using a typical volcanic crustal model including a stochastic component. We locate the earthquakes using both an average one-dimensional crustal model and a three-dimensional model of the kind that local-earthquake tomography produces. Finally, we simulate ray-path travel time corrections of the kind expected from a calibration experiment, and compare the locations obtained using those corrections with those obtained using only the three-dimensional tomographic model.