Assessing the benefit of 3D a priori models for earthquake location

Abstract:

Earthquake location in 1D Earth models is a routine procedure. Particularly in environments such as subduction zones where the network geometry is biased and lateral velocity variations are large, the use of a 1D model can lead to strongly biased solutions. This is well-known and it is therefore usually preferred to use three-dimensional models, e.g. from local earthquake tomography. Efficient codes for earthquake location in 3D models are available for routine use, for example NonLinLoc. However, tomographic studies are time-consuming to carry out, and a sufficient number of data might not always be available. However, in many cases, information about the three-dimensional velocity structure is available in the form of refraction surveys or other constraints such as gravity or receiver functions based models. Failing that, global or regional scale crustal models could be employed. However, it is not obvious that models derived using different types of data lead to better location results than an optimised 1D velocity model. On the other hand, correct interpretation of seismicity patterns often requires comparison and exaxt positioning in pre-existing velocity models. In this presentation we draw on examples from the Chilean and Sumatran margins as well as a mid-ocean ridge environments, using both data and synthetic examples to investigate under what conditions the use of a priori 3D models is expected to result in improved location results and modifies interpretation. Furthermore, we introduce MATLAB tools that facilitate the creation of three-dimensional models suitable for earthquake location from refraction profiles, CRUST1 and SLAB1.0 and other model types.