Hans Israelsson

Validation of Regional and Teleseismic Travel-Time Models by Relocating Ground-Truth Events

Recently developed 3D global seismic velocity models have demon- strated location improvements through independent regional and teleseismic travel- time calibration. Concurrently, a large set of high-quality ground-truth (GT) events with location accuracies 10 km or better (GT0-GT10) has been collected for Europe, the Mediterranean, North Africa, the Middle East, and western Eurasia. In this study we demonstrate event location improvements using this new data set by applying the regional and teleseismic model-based travel-time calibrations (independently and jointly). Besides relocating events using all arrivals, a subset of the GT events was also relocated using controlled station geometries generated from a constrained boot- strapping technique. This approach simulates sparse networks and reduces the effect of correlated errors to ensure valid 90% error ellipse coverage statistics. With respect to the GT events, we compared event relocations, with and without travel-time cal- ibrations, considering statistics of mislocation, error ellipse area, 90% coverage, or- igin time bias, origin time errors, and misfit. Relocations of over 1200 GT events show that Pn and/or P calibration reduced mislocation for 60%-70% of the events. Joint regional Pn and teleseismic P travel-time calibration provided the largest lo- cation improvements and achieved approximately GT5 accuracy levels. Due to cor- related errors, event locations using large numbers of stations have deficient 90% error ellipse coverage. However, the coverages derived from the model errors are appropriate for the simulated sparse regional and teleseismic networks. Our valida- tion effort demonstrates that the global model-based calibrations of Pn and teleseis- mic P travel times reduce both location bias and uncertainty over wide areas.

Collection of a Reference Event Set for Regional and Teleseismic Location Calibration

A three-year consortium project, with members of Science Applications International Corp., University of Colorado at Boulder, Harvard University, Multimax Inc., Geophysical Institute of Israel, Western Services, and University of California at San Diego, was initiated in 2000 to improve locations and reduce uncertainties in the Middle East, North Africa, Europe, and Western Eurasia. The consortium developed high-resolution three-dimensional models of the Earths mantle to generate accurate travel-time predictions for regional and teleseismic P phases. Since the approach was purely model-based, a large set of high-quality reference events was needed to validate the model predictions. The consortium has spent considerable effort to collect, vet, and validate reference events located with 5-km accuracy or better by local networks and “promoted” reference events located with an accuracy of 7 km or better by application of multiple-event location techniques. Consortium members built an extensive network of contacts to solicit candidate reference events from local, regional, and national network operators. Strict methodologies were developed to identify candidate reference events in earthquake bulletins, and to validate and quality control the selected candidate reference events. The outcome of the consortium effort was a quality-controlled reference event list with nearly 2000 events and over 200,000 arrivals. The Reference Event List is provided as an electronic supplement to this article. Online material : Reference event database. Manuscript received 23 June 2003.

Optimum subarray configuration using genetic algorithms

Subarray configuration is not a trivial problem in array signal processing. A proper subarray configuration is important to improve the detectability of an array. A new searching algorithm, which is based on genetic algorithms (GA), for the optimum subarray configuration is proposed. Our preliminary application to a seismic array has indicated that the new algorithm can search a population of subarrays in a more efficient and robust way. The beamforming gain of the optimum subarray derived by the GA is very close to the theoretical gain. Experimental results on signal detections have demonstrated that a beamforming recipe with optimum subarrays can provide further enhanced signal-to-noise ratio (SNR), compared to a recipe without a subarray configuration. The approach proposed here can be easily extended to the weight determination problem for the weighted beamforming process by using multi-bit instead of 1-bit representation for each sensor in the chromosome model.