Reliability of earthquake location procedures in heterogeneous areas: synthetic tests in the South Western Alps, Italy

The accuracy of procedures for locating earthquakes and tomographically ima ging 3-D velocity structures, is influenced by factors such as network geom etry, errors in picking seismic phases, modeling errors and velocity model uncertainties, The main purpose of this study is to investigate the importa nce of different error sources through synthetic experiments. The synthetic data sets were generated for a realistic 3-D velocity model o f the South Western Alps adopting a precise raytracing technique. The actua l station configuration and recorded seismicity during the passive seismolo gical experiment Geofrance3D-Alps, conducted between August 1996 and Februa ry 1997, were used. The station configuration corresponded to an enhancemen t of the regional networks operating in the area, with the addition of temp orary stations, In addition, synthetic data sets were corrupted by applying different error schemes. Furthermore, to assess the locating capabilities of the regional networks, synthetic data sets were constructed for syntheti c events placed in correspondence to the nodes of a regular three dimension al (3-D) grid. We mimicked the actual data management scheme for synthetic data processing : we re-located the earthquakes, performed an inversion for a minimum 1-D v elocity model with station corrections and, then, ran a complete 3-D veloci ty model inversion. One of the main results, confirming earlier studies was that earthquake loc ations in strongly heterogeneous areas, obtained using a regional 1-D model , can have significant errors. Even a minimum 1-D model with station correc tions does not significantly improve the accuracy of parameter locations in the South Western Alps. For high precision locations, a 3-D model is neede d. However, a 3-D model cannot be reconstructed from noise contaminated dat a recorded by a sparse regional network, (C) 2001 Elsevier Science B.V, All rights reserved.