Distributed Multigrid Technique for Seismic Tomography in Sensor Networks

Abstract

The modern seismic sensor used to monitor volcanic activity can record raw seismic data for several years. This massive amounts of raw data recorded consists of information regarding earthquake’s origin time, location, velocity the wave traveled etc. To extract this information from the raw samples, current state of the art volcano monitoring systems rely on gathering these high volume data back to a centralized base station. From these extracted information vulcanologist are able to compute tomography inversion and image reconstruction to understand the magma structure beneath. These high volume data are mostly gathered manually or sometimes relayed using powerful expensive broadband stations making vulcanologist unable to obtain real time information of the magma structure and also to predict the occurrence of eruption. Also, the sheer volume of raw seismic data restricts the deployment of large numbers of seismic sensors over the volcano making it difficult to obtain high resolution imagery. To overcome these challenges, a new in-network distributed method is required that can obtain a high resolution seismic tomography in real time. In this paper, we present a distributed multigrid solution to invert seismic tomography over large dense networks, performing in-network computation on huge seismic samples while avoiding centralized computation and expensive data collection. This new method accelerates convergence, thereby reducing the number of message exchanges required over the network while balancing the computation load (Our research is partially supported by NSF-CNS-1066391, NSF-CNS-0914371, NSF-CPS-1135814 and NSF-CDI-1125165.)