Geophysics Seminar: How leaky is the leaky Dead Sea transform?

Abstract: 

The problem of learning from seismic recordings plays a central role in many applications like ongoing detection of natural seismic events, discrimination between natural earthquakes and man-made explosion and in the last decade, for monitoring micro-seismic activities that result from oil and gas drilling.  There is a growing interest of developing automatic mechanisms for modeling and identifying the properties of a seismic event, in order to assist the analysts in their routine, partly manual, processing. One main motivation that is addresses in this work is the ability have a reliable identification of manmade explosions, and separate them from the natural events.  In the proposed framework, the raw seismic waveforms are pre-proceed and represented in the time-frequency domain, then, kernel based and kernel fusion machine learning methods are applied to model the data into a low-dimensional space, in which analysis is carried out.  The proposed method is tested several datasets that were recorded in Israel and in Jordan. The method achieves promising results in classification of event type as well as in estimating the location of the event. 
Joint work with Dr. Yuri Bregman and Dr. Yochai Ben-Horin from Soreq Nuclear Research Center, Israel, Dr. Ofir Lindenbaum, Yale University and Prof. Amir Averbuch, Tel-Aviv University.

Seminar Organizers: Prof. Moshe Reshef and Dr. Alon Ziv