LLNL's REGIONAL SEISMIC DISCRIMINATION RESEARCH

As part of the Department of Energy's research and development effort to improve the monitoring capability of the planned Comprehensive Nuclear-Test-Ban Treaty international monitoring system, Lawrence Livermore Laboratory (LLNL) is testing and calibrating regional seismic discrimination algorithms in the Middle East, North Africa and Western Former Soviet Union. The calibration process consists of a number of steps: 1) populating the database with independently identified regional events; 2) developing regional boundaries and pre-identifying severe regional phase blockage zones; 3) measuring and calibrating coda based magnitude scales; 4a) measuring regional amplitudes and making magnitude and distance amplitude corrections (MDAC); 4b) applying the DOE modified kriging methodology to MDAC results using the regionalized background model; 5) determining the thresholds of detectability of regional phases as a function of phase type and frequency; 6) evaluating regional phase discriminant performance both singly and in combination; 7) combining steps 1-6 to create a calibrated discrimination surface for each stations; 8) assessing progress and iterating. We have now developed this calibration procedure to the point where it is fairly straightforward to apply earthquake-explosion discrimination in regions with ample empirical data. Several of the steps outlined above are discussed in greater detail in other DOE papers in this volume or in recent publications. Here we emphasize the results of the above process: station correction surfaces and their improvement to discrimination results compared with simpler calibration methods. Some of the outstanding discrimination research issues involve cases in which there is little or no empirical data. For example in many cases there is no regional nuclear explosion data at IMS stations or nearby surrogates. We have taken two approaches to this problem, first finding and using mining explosion data when available, and second using test-site based models to transport earthquake-explosion discrimination behavior to new regions. Finally an important component of our research is assessing improvement in the ability to discriminate events. By combining the multivariate discriminants with the threshold detection curves for the regional seismic phases used in those discriminants, we have started to make maps of the probability an event will be identified properly. These maps serve a broad range of purposes from demonstrating progress to funding agencies to prioritizing research and calibration efforts.