Location of seismic events using inaccurate data from very sparse networks

Abstract

Seismological experiments on planets other than Earth are difficult to install and generally very expensive. Additionally, the placement of these stations is determined much more by engineering and accessibility constraints than on Earth. It is thus likely that the resulting networks consist of few stations only and do not have an ideal geometry for location purposes. This study presents a new location scheme which aims towards a description of the entire solution set of the location problem, which is a non-linear inversion problem. The assumptions on the underlying error statistics are reduced from the usual assumption of a certain distribution to assuming a finite width of the distribution only, but without referring to a certain functional expression. This allows to formulate a falsification criterion, which tests for each individual observation if it is compatible with a certain candidate location or not. A deterministic grid search, based on a locally adaptive triangle grid derived from an initial icosahedron, focuses the search to those portions of the planetary interior that are compatible with most observations, and eventually returns a set of candidate hypocentres that are not falsified by any arrival time or back azimuth data. The implementation of this method is tested with synthetic and real data examples and proves to be able to handle complexly shaped and disconnected sets of possible solutions.