Probabilistic Hypocenter-Velocity Determination for Moderate Local Earthquakes Using a Sparse Network

Abstract:

The most widely used algorithms to locate earthquakes are based on iterative linearized least-squares techniques. The drawbacks of theses methods, in particular when only a small amount of observations is available, are large dependencies on location and origine time due to a priori assumptions on velocity models and on an initial set of hypocentral parameters. However, the determination of accurate earthquake hypocentral parameters with estimation statistics is essential to constrain seismic structures and to understand ongoing seismotectonic processes.

We develop a joint hypocenter-velocity determination method using Bayesian inferences and apply it to data recorded within a sparse network of stations. Monte Carlo sampling methods, using Markov chains, generate models within a broad range of continuous structure and hypocenter parameters, distributed according to the unknown posterior distribution. The exploration for velocity structure is defined by flat homogeneous layers with velocity and thickness parameters. Output covariances on parameters characterize the trade-off between unknowns, e.g. hypocentre depth and origin time. This method is useful to reduce uncertainties on hypocentral parameters in the absence of a priori solution for regions in which the velocity structure is poorly constrained. The algorithm has been successfully used to accurately locate events of the Armorican Massif, an extensive outcrop of Variscan basement in western France, with moderate and diffuse local seismicity.