Fast Algorithm for Continuous Monitoring with Ambient Noise

A common approach to analyzing ambient seismic noise involves O(n^2) pairwise cross-correlations of n sensors. Following cross-correlations the resulting coherent waveforms are then synthesized into a velocity estimate, often in the form of a dispersion image. As we move towards larger surveys and arrays for continuous subsurface monitoring, this computation can become prohibitively expensive. We show that theoretically equivalent results can be achieved by a simple algorithm which skips the cross-correlations, and scales as O(n). Additionally, this algorithm is embarrassingly parallel, and is significantly cheaper than the commonly used algorithms. We demonstrate the algorithm on two field data sets: (1) a continuously recording linear trenched distributed acoustic sensing (DAS) array designed as a pilot test to develop a permafrost thaw monitoring system, and (2) the Long Beach Array, an irregularly spaced 3D array. These results show superior performance in both speed and numerical accuracy. An open-source implementation of this algorithm is available.