Real-Time Cooperative Analytics for Ambient Noise Tomography in Sensor Networks

Abstract

The transformative integration of sensor networks and geophysical imaging techniques enables the creation of a system to monitor and analyze seismic data in real time as well as image various subsurface structures, properties, and dynamics. Ambient noise seismic imaging is a technique widely used in geophysical exploration for investigating subsurface structures using recorded background raw ambient noise data. The current state-of-the-art of ambient noise monitoring relies on gathering these high volumes of raw data back to a centralized server or base station to pre-process, cross-correlate, analyze frequency-time components, and generate subsurface tomography. However, modern computational sensors (for example, those with $\sim$ 1.2 GHz of processor and $\sim$ 1 GB of memory) can be not only used for recording raw vibration data but also performing in situ processing and cooperative computing to generate subsurface imaging in real time. In this paper, we present a distributed solution to apply ambient noise tomography over large dense networks and perform in-network computing on huge seismic samples while avoiding centralized computation and expensive data collection. Results show that our approach can detect subsurface velocity variations in real time while meeting network bandwidth constraints and reducing communication cost ( $\sim -75\%$ ).