Analysis of instrument self-noise and microseismic event detection using power spectral density estimates

Abstract

Reliability of microseismic interpretations is very much dependent on how robustly microseismic events are detected and picked. Various event detection algorithms are available but detection of weak events is a common challenge. Apart from the event magnitude, hypocentral distance, and background noise level, the instrument self-noise can also act as a major constraint for the detection of weak microseismic events in particular for borehole deployments in quiet environments such as below 1.5-2 km depths. Instrument self-noise levels that are comparable or above background noise levels may not only complicate detection of weak events at larger distances but also challenge methods such as seismic interferometry which aim at analysis of coherent features in ambient noise wavefields to reveal subsurface structure. In this paper, we use power spectral densities to estimate the instrument self-noise for a borehole data set acquired during a hydraulic fracturing stimulation using modified 4.5-Hz geophones. We analyse temporal changes in recorded noise levels and their time-frequency variations for borehole and surface sensors and conclude that instrument noise is a limiting factor in the borehole setting, impeding successful event detection. Next we suggest that the variations of the spectral powers in a time-frequency representation can be used as a new criterion for event detection. Compared to the common short-time average/long-time average method, our suggested approach requires a similar number of parameters but with more flexibility in their choice. It detects small events with anomalous spectral powers with respect to an estimated background noise spectrum with the added advantage that no bandpass filtering is required prior to event detection. © 2014 The Authors Published by Oxford University Press on behalf of the Royal Astronomical Society.