Accession Number : ADA462267


Title :   Improved Event Location Uncertainty Estimates


Descriptive Note : Conference paper


Corporate Author : SCIENCE APPLICATIONS INTERNATIONAL CORP SAN DIEGO CA


Personal Author(s) : Bondar, Istvan ; McLaughlin, Keith ; Israelsson, Hans


Full Text : https://apps.dtic.mil/dtic/tr/fulltext/u2/a462267.pdf


Report Date : 21 Sep 2006


Pagination or Media Count : 10


Abstract : While many recent studies aimed to reduce location bias by introducing improved travel-time corrections, less effort was devoted to the complete estimation of location uncertainty, despite the fact that formal error ellipses are often overly optimistic. Since most location algorithms assume that the observations are independent, correlated systematic errors that are due to similar ray paths inevitably result in underestimated location uncertainties. Furthermore, the tails of real seismic data distributions are heavier than Gaussian. The main objectives of this project are to develop, test, and validate methodologies to estimate location uncertainties in the presence of correlated, systematic and non-Gaussian errors. Particular attention is paid to robust and transportable models for a travel-time covariance matrix. To address correlated errors, we estimate the spatial correlation structure in arrival-time data using variogram models. We developed a methodology based on copula theory to derive robust, data-driven variogram models. For validation purposes, we use GT0-2 event clusters. These include the Nevada Lop Nor, Semipalatinsk, and Novaya Zemlys test sites, as well as the Azgir Peaceful Nuclear Explosions and the Lubin Poland, mine-related events. Using ground-truth (GT) clusters allows us to calculate ground truth residuals with respect to the GT locations for a specific velocity model. We show the improvements in variogram estimates when using global 3D models instead of the IASP91 model. The proper choice of the underlying velocity model is especially important for regional phases. To address issues embodied by real data distributions, we performed fully controlled experiments using known high signal-to-noise ratio (SNR) waveforms, scaled down to several magnitude levels and embedded in clean noise to derive models of measurement errors.


Descriptors :   *ERROR ANALYSIS , *POSITION(LOCATION) , SEISMIC DATA , COVARIANCE , SYMPOSIA , SIGNAL TO NOISE RATIO


Subject Categories : Seismology
      Statistics and Probability
      Cybernetics
      Seismic Detection and Detectors


Distribution Statement : APPROVED FOR PUBLIC RELEASE