Epicentre accuracy based on seismic network criteria

Abstract

We establish reliable and conservative estimates for epicentre location accuracy using data that are readily available in published seismic bulletins. A large variety of seismic studies rely on catalogues of event locations, making proper assessment of location uncertainty critical. Event location and uncertainty parameters in most global, regional and national earthquake catalogues are obtained from traditional linearized inversion methods using a 1-D Earth model to predict traveltimes. Reported catalogue uncertainties are based on the assumption that error processes are Gaussian, zero mean and uncorrelated. Unfortunately, these assumptions are commonly violated, leading to the underestimation of true location uncertainty, especially at high confidence levels. We find that catalogue location accuracy is most reliably estimated by station geometry. We make use of two explosions with exactly known epicentres to develop local network location (0°–2.5°) accuracy criteria. Using Monte Carlo simulations of network geometry, we find that local network locations are accurate to within 5 km with a 95 per cent confidence level when the network meets the following criteria: (1) there are 10 or more stations, all within 250 km, (2) an azimuthal gap of less than 110°, (3) a secondary azimuthal gap of less than 160° and (4) at least one station within 30 km. To derive location accuracy criteria for near-regional (2.5°–10°), regional (2.5°–20°) and teleseismic (28°–91°) networks, we use a large data set of exceptionally well-located earthquakes and nuclear explosions. Beyond local distances, we find that the secondary azimuthal gap is sufficient to constrain epicentre accuracy, and location error increases when the secondary azimuthal gap exceeds 120°. When station coverage meets the criterion of a secondary azimuth gap of less than 120°, near-regional networks provide 20 km accuracy at the 90 per cent confidence level, while regional and teleseismic networks provide 25 km accuracy at the 90 per cent confidence level.